Semantic Web Interest Group IRC Chat Logs for 2001-11-21

01:27:52 * danja reads timbl's next novel in the logs...

01:30:09 <danja> "terrible code"?

01:31:20 <danja> "munging"?

01:35:53 <danja> "... without need for an ivory tower"

01:36:00 <danja> heh!

01:36:07 <danja> thanks Tim

01:36:28 <danja> it looks good though

01:36:54 <danja> hendry 101?

01:37:52 <danja>http://www.isacat.net/svg/showme.html

01:37:52 <dc_rdfig> A: http://www.isacat.net/svg/showme.html from danja

01:38:04 <danja> A: SVG is pish

01:38:04 <dc_rdfig> commented item A

01:38:25 <danja> A: can even do it on MS servers...

01:38:25 <dc_rdfig> commented item A

01:42:35 <danja> A: hey Aaron, do you run servlets?

01:42:35 <dc_rdfig> commented item A

01:42:48 <AaronSw> I'm not running any now

01:43:10 <danja> heh - thought you'd be there

01:43:46 <danja> what are you working on now?

01:45:58 <AaronSw> reverse engineering Google, at the moment

01:46:11 <danja> what, backlinks?

01:46:28 <danja> or distribution?

01:46:48 <AaronSw> too busy to explain, sorry

01:50:50 <AaronSw> ask me later

01:51:02 <danja> any bots listening - Norvig:foaf(or at least recommendation):dan_c:works-for:google

01:52:09 <danja> ok Aaron, got a cat cat to munge myself...

01:52:19 <AaronSw> ehm ok

01:53:09 <sbp> erm...

02:30:08 * AaronSw disconnects

02:33:29 * AaronSw reconnects

02:37:27 <DanC> re mime type major/minor: that's not a case for peeking in URIs; to model that with URIs, you need 2 URIs (the minor type and the major type) and an RDF statement connecting them.

02:38:24 <DanC> so text/plain is a pair (http://iana.org/media-types/text http://iana.org/media-types/text/plan) plus an arc { <...text/plain> mime:subTypeOf <.../text> }.

02:39:24 <DanC> in practice, you'd just distribute the one <text/plain> URI (with an implicit base of http://iana.org/...) and count on everybody having the knowledge that it's a subtype of text/ cached locally

02:39:45 <DanC> for lesser-known URI schemes, they'd have to look up the base type in the web

04:00:04 * sandro wants an ODBC driver for RDF. It would be good. RDF over ODBC would also be good, but ODBC over RDF would let Joe Random database user (eg my sister using Cold Fusion) get ahold of the universe of RDF data.

04:00:39 <DanC> hmm... interesting

04:01:34 <DanC> tricky bit would be condensing (a) a list of URIs from which to slurp RDF and (b) a mapping of their schema(s) into an SQL database into an ODBC "data source"

04:02:05 <DanC> but I think Squish exposes an ODBC interface, no?

04:03:07 <DanC> sandro, wanna educate me about PC hardware a little bit? e.g. is it OK to buy a storebrand PC, or should I care about the motherboard, cache, that sort of thing?

04:04:59 <sandro> Squish appears to be ODBC-like, but the syntax is not really SQL, so you can't use it from, eg Cold Fusion. We need the syntax to be exact, so you don't know if you're using Oracle, SQLServer, or RDF, once you establish (point to?) the mapping from RDF identifiers into SQL identifiers. Should that mapping be on the web, or just local to your process? I think local.

04:05:13 <DanC> ah.

04:05:39 <DanC> local/web: I would hope that's an option.

04:06:02 <DanC> i.e. I'd like to be able to share it once I've got it figured out. And I sure as hell want to be able to use somebody else's once they've figured it out.

04:07:04 <sandro> Re: PC hardware -- Um, you probably should care a bit, esp if you're going to run linux. I usually try to read a few reviews of each part being used with linux before I buy.

04:07:59 <DanC> ok, well, I'm trying to learn how to read said specs.

04:08:38 <DanC> e.g. "Abit (KT7E) MB w/AMD K7 (1.2G

04:08:38 <DanC> Hz 200Bus)w/Monster Fan Fedex

04:08:38 <DanC> 2Days"

04:08:38 <sandro> I'd expect the mapping to be mostly saying: rdf identifiers starting with "http://www.w3.org/1999/02/22-rdf-syntax-ns#" should map to/from local identifiers starting with "_rdf_" or some such (I'm not sure what identifier characters are legal in SQL table and column names)

04:09:24 <DanC> and "Matrox G450 Dual Head 32MB AG

04:09:24 <DanC> P Video Card (OEM)"

04:09:59 <DanC> I invested in SCSI last time around. I think that was a waste. I

04:10:26 <DanC> I'm still bothered by the idea that every byte off the disk would go thru the CPU, but I gather with "ultra-DMA", that's no longer the case. But I'm not sure.

04:12:19 * DanC is reading up on PC disk technology in http://thef-nym.sci.kun.nl/cgi-pieterh/atazip/atafq-2.html

04:13:18 <sandro> Yeah - spend the money on RAM instead of SCSI. There's no reason not to have 1.5G of RAM, and you wont be going to the disk much with that kind of cache.

04:13:55 <DanC> 1.5G of ram: it's easy to buy a motherboard that won't fit that much RAM, no?

04:14:43 <sandro> I think most of the modern K7 boards (like the KT7E) will take three modules, and 512M modules are ~50 I think.

04:14:56 * DanC remembers working on his first 1GB machine; it was a supercomputer shared with around 30 other users in the early '09s

04:14:58 <DanC> '90s

04:15:18 <sandro> I know, it's nuts. :-)

04:15:31 <DanC> speak slowly, please. "modern K7 board"? I didn't know there were.. uh... classical K7 boards.

04:16:16 <sandro> Yeah $45 for 512MF of PC150. If you you have the budget for a DDR machine it's $79. (pricewatch)

04:16:32 <sandro> s/MF/MB

04:16:44 * DanC is also learning about Mac hardware in #openmac. These claims that Mac hardware is price-competitive seem out in the weeds. Cheapest G4 mac is $1600 or so, and a screamin linux-PC can be had for under $1000.

04:17:50 <sandro> I think K7 boards cycle every 6 months or a year, as the clock speeds go up and new motherboard chipsets come out.

04:18:01 <sandro> (same with Intel-CPU boards)

04:18:12 <AaronSw> DanC, have you looked at refurbished models?

04:18:26 <DanC> ah! "K7 board" means "motherboard for a K7 processor".

04:19:11 <DanC> no, AaronSw, I haven't looked at refurbished models. Why should I? Again, price-competitiveness claims would seem to be out in the weeds. re-furbished PCs can be had for like 50 cents.

04:19:22 <AaronSw> Heh.

04:20:02 * DanC tries to remember what a K7 processor is

04:20:42 <sandro> It's an Athlon

04:20:53 <DanC> ah yes.

04:21:34 <DanC> there's no reason to choose Intel over Althon or whatever, as long as I'm running linux, right? I mean: none of the linux stuff knows the difference, true?

04:22:07 <DanC> ... which means: MPEG video performance is likely to suck, if I ever needed to do that.

04:22:55 * DanC is frightened by BIOS compatibility issues in this PC disk FAQ.

04:23:19 <sandro> Right. I'm an Athlon fan these days, for 20% technical and 80% political reasons. :-)

04:23:34 <DanC> sandro, BIOS comes with the motherboard, right? Is there a significant risk of buying a motherboard with BIOS problems?

04:24:33 <DanC> how about video hardware, sandro? What is "AGP" anyway? do I care? how the hell do I choose a graphics card?

04:24:43 <sandro> I've never heard of that. As I understand it, linux doesn't use the bios once the kernel's running.

04:25:25 <sandro> AGP is the current standard for graphics card busses. They have their own slot, the AGP slot.

04:26:13 <DanC> ah... so AGP is a physical form-factor deely, like PCI vs. [the other one that I can't remember]

04:26:26 <sandro> ANy card more than $50 should be fine -- matrox has a reputation for higher image quality, so I bought matrox. Also, dual head might be cool. I got the card, but I've never bothered to plug in my second monitor and configure X for it.

04:27:12 <sandro> Yep. VESA was the video slot for ISA machines, AGP is the video slot for PCI machines.

04:27:35 <sandro> (once they figured out ISA and PCI weren't really good enough for video.)

04:27:51 <DanC> hm... I've spent enough time cursing at XF86Config files to wonder about that "any card over $50 should be fine" claim.

04:28:06 <AaronSw> Macs support dual-head by default.

04:28:17 <sandro> If you actually want spiffy video games of course you want a GeForce, but I bet that's not a big factor for you.

04:28:23 <AaronSw> See, I think it's little things like this that make the "off into the weeds" claim not so crazy

04:28:53 * mnot is seriously considering switching from Linux to MacOS X

04:29:00 <DanC> well, for the price of a Mac, I could hire a linux weeny to do all this for me. And a masseur to pass the time too.

04:29:14 <sandro> Okay, maybe not ANY card. But the field has gotten very narrow -- almost all cards use the same chipsets.

04:29:27 <AaronSw> Heh @ massuer

04:29:47 <AaronSw> I didn't mean configuration, I meant stuff like higher-quality graphics cards, etc.

04:29:59 * DanC is still seriously considering a Mac too, but is pretty annoyed at the prices.

04:30:17 <mnot> It's not the setup, it's the ongoing tweaking that I'm finding isn't worth it (after 7 years on Linux and Slolaris... argh)

04:30:22 <mnot> iBook ;)

04:30:49 <bijan> 5% of the market...can't expect equivalent prices ;)

04:30:55 <sandro> DanC, I sent you e-mail with my system config (purchased in July). I still love it, though there is some glitch that makes Windows crash on the rare occasions I boot it. (My sister bought a pre-built system from the same people, and it had a glitch, and she had trouble returning it (it was being shipped to NYC during Sept-11), and ended up just returning it and paying the 15% restocking fee.)

04:31:02 <DanC> hmm... I just bought a laptop; they're gonna look at me funny if I ask for an ibook, I'm afraid. But I suppose I could get away with it.

04:31:25 <DanC> sandro, I have your system config from July bookmarked. You may note I quoted from it above ;-)

04:31:27 <mnot> or just go for an iMac, as long as the screen is big enough.

04:31:28 <AaronSw> Yeah, the Mac portables make better desktops than their desktops do, IMO

04:32:05 <DanC> would an ibook with an external USB keyboard and 17" flat-screen monitor work well?

04:32:11 <sandro> Ah, DanC, no I didn't notice the direct quote, though it did look very familiar. :-)

04:32:28 <AaronSw> http://dealmac.com/sections/systems.html might be of some help

04:32:34 <AaronSw> We bought th $400 iMac and it's awesome

04:32:49 <AaronSw> I don't think the iBook goes well with external monitors

04:33:08 <mnot> If you get the iBook, make sure to get the 600Mhz - it's got a 100Mhz bus

04:33:36 <DanC> "the $400 iMac"? I don't see it

04:33:38 <bijan> aaron: Why? I think it does mirroring.

04:33:51 <AaronSw> oh... right

04:34:04 <AaronSw> I was thinking DanC wanted dual-head for some reason...

04:34:06 <bijan> Admittedly, I was peeved when they stopped doing dual screens with the laptops.

04:34:22 <AaronSw> Err, $500 iMac, DanC, oops.

04:34:56 <mnot> it would be *really* nice if they had a docking station for the iBook or powerbook...

04:34:57 <DanC> this one? http://dealmac.com/articles/28356.html

04:35:11 <AaronSw> yeah

04:35:18 <AaronSw> altho i'm not sure if it meets your needs

04:35:38 <DanC> er... did you drive to NY to get it?

04:36:09 <AaronSw> no, they shipped it to us

04:36:14 <DanC> oh.

04:36:32 <DanC> you're not running OS-X on it, are you?

04:36:49 <bijan> Ooo, 20" Trinitron, refurb $149

04:37:38 <AaronSw> not right now, but it can boot into OS X

04:38:33 <DanC> er... I pretty much depend on emacs for software development. Does emacs run in OS9?

04:38:44 <AaronSw> Yep.

04:38:54 <AaronSw> But you're probably better off with OS X

04:38:55 * DanC might like to stop using emacs, actually

04:39:11 <AaronSw> I found I became much more productive with a more GUI editor

04:39:21 <AaronSw> but i never learned emacs key-commands...

04:39:24 <DanC> I just want ONE cut/copy/paste mechanism EVERYWHERE.

04:39:55 <bijan> Indeed.

04:40:00 <DanC> I CANNOT STAND HAVING TO *THINK* ABOUT WHICH KEYS TO USE FOR CUT/COPY/PASTE IN WHICH APPS.

04:40:07 <DanC> er... ahem. Sorry.

04:40:13 <AaronSw> I've never used anything else... altho it is sort of annoying to delete emacs \s...

04:40:15 <bijan> Heh.

04:41:06 <bijan> You think that's bad. Squeak tries to be consistent accross platforms. So on windows in squeak you have alt-c/v/x, and out of squeak ctrl-c/v/x. now *that* sucks.

04:41:25 <AaronSw> ouch

04:42:11 <DanC> that's the truely annoying thing about Amaya... it's the only app in the world that groks cut/copy/paste of hypertext... ALMOST. It tempts you into thinking it knows what you want it to do, and then just after you start to trust it, BLAMMO!

04:42:29 <AaronSw> Hi Mark!

04:42:37 <Mark> Hey Aaron

04:42:37 <bijan> Heh. that's not the only place it lures you in to stomp on your head.

04:42:38 <mnot> Yeah. Hopefully, the GTK port will mature

04:42:56 <DanC> the GTK port rocks, yes.

04:43:02 <Mark> Geez, the gang's all here. Hey Mark.

04:43:31 <mnot> hey Mark

04:43:40 <DanC> topic should be more like "life, PC vs Mac, and everything"

04:43:57 <AaronSw> :-)

04:44:01 <mnot> hopefully we won't drift into emacs vs vi

04:44:08 <bijan> When I was recommending Amaya and people asked me about its stability I kept saying, "It's really really good at not..umh...loosing work when..er...it goes down hard as it sometimes does a lot."

04:44:27 <bijan> this pitch needed work :)

04:44:28 <AaronSw> Heh.

04:44:47 <DanC> LOL!!!!!

04:45:04 <AaronSw> Better to fix the code than to fix the pitch ;-)

04:45:12 * DanC is keeping a page dedicate to the persuit of the PerfectOffice http://dm93.org/y2001/po23

04:45:27 <AaronSw> why the y2001?

04:45:34 * DanC should make it a wiki, but has been too lazy thus far; no excuse when dm93.org is zope-powered, too.

04:45:34 <AaronSw> Zope doesn't like directories that start w/ numbers?

04:45:51 <DanC> no, I mapped /2001/ to the filesystem.

04:46:08 <AaronSw> hmm?

04:46:27 <AaronSw> oh, I see

04:46:40 <DanC> when I started dm93.org, it was apache-powered.

04:46:49 <bijan> "Well, considering that various versions of Word would completely destroy your file upon *saving* it, and, nowadays, might well reformat your hard drive while mailing copies of itself to everyone in your address book, I think preserving work in the face

04:47:04 <DanC> I took my apache /2001/ files and exported them thru Zope's legacy bow to the filesystem

04:47:22 <bijan> of, uhm, hard crashes that..well..if you do the wrong thing can happen every thirty seconds...ah...is a *commendable* quality."

04:47:33 <AaronSw> Heh.

04:47:42 <bijan> "Did I mention it's open source?"

04:48:09 * DanC isn't sure whether bijan's doing advocacy or a stand-up bit ;-)

04:49:02 * bijan is an iffy marketer at best.

04:49:15 <DanC> what brings you by these parts, Mark?

04:50:04 <Mark> Just did an ego surf on "RESTwiki". Found your log. neat stuff.

04:50:31 <DanC> ah yes... backlinks. the 8th wonder of the world, that ;-)

04:51:19 <DanC> RESTwiki seems to be having impact. I gather the W3C XML Protocol is going to do The Right Thing w.r.t. 200 vs. 500, if not GET vs. POST.

04:51:43 <mnot> Don't think GET vs. POST was ever logged as an issue

04:51:52 <DanC> the hell it wasn't.

04:51:59 <Mark> I'm not sure the wiki had much to do with that, though that would be nice if it did

04:52:01 * sandro wanders off to sleep (hopefully)

04:52:06 <DanC> I sat on Hugo's desk while he did it, practically ;-)

04:52:15 <AaronSw> Heh!

04:52:22 <mnot> hmm

04:52:31 <DanC> well, maybe not RESTwiki, but the pro-REST argument in the XML P WG in general.

04:52:36 <mnot> Most of those issues were discussed at the Sept f2f, but it wasn't

04:53:02 <Mark> Yah, perhaps. Henrik didn't do any REST preaching, but I think his influence was the main reason

04:53:13 <mnot> definately.

04:53:15 <Mark> He changed SOAP 1.0 from 200 to 500 after all

04:53:23 <Mark> (in 1.1)

04:53:27 <DanC> ah.

04:53:40 <mnot> He was the strongest voice for it

04:53:48 <mnot> of course, he was also arguing for SOAPAction ;)

04:53:50 <DanC> I was going to say: isn't HFN an author of the rather REST-unfriendly earlier versions of SOAP? I'm sorta confused about that.

04:54:17 <Mark> Not to my knowledge. He came on to clean up SOAP 1.0.

04:54:25 <DanC> interesting.

04:54:38 <Mark> .. when he got redirected to MS on his way to UCI 8-)

04:54:53 * AaronSw makes mental note not to go to UCI w/o strong protection.

04:54:54 <mnot> yeah, he doesn't show as an author of the nov 199 I-D

04:55:11 <DanC> Aaron, speaking of small-world/backlink stuff, that perl implementation of N3, and the discussion at scripting.foo, blew me away.

04:55:29 <AaronSw> Yeah, I was quite surprised at that.

04:55:29 <Mark> IIRC, he went to MS in the summer of 99

04:55:50 <DanC> yes, AaronSw/Skywalker, the force is strong in those parts. Beware.

04:55:56 <AaronSw> Heh.

04:56:03 <AaronSw> Dave and I had a very nice chat and I think we worked over a lot of his concerns w/ RDF

04:56:15 <AaronSw> Wouldn't it be the Dark Side?

04:57:08 <AaronSw> Did you email the Perl N3 guy?

04:57:17 <AaronSw> I'm interested to hear where he came from

04:57:25 <DanC> N3 is a fascinating phenomenon, to me. I told TimBL he was out of his mind when he first wrote the design note. Then I implemented it ;-) EricM is all worried it's gonna take the wind out of RDF/xml, but it's winning more converts than any other RDF tool that I can think of.

04:57:47 <DanC> no, haven't opened a channel to the Perl N3 guy.

04:57:59 <Mark> I'm gonna hit the hay, but I'll be back.

04:58:04 <DanC> ciao

04:58:08 <AaronSw> c'ya

04:58:15 <Mark> btw, I just read up on N3 yesterday. very nice.

04:58:30 <Mark> later

04:58:54 <AaronSw> Hmm, another Aaron seems to be interested in RDF::N3 -- http://aaronland.info/weblog/category/181/recent

04:59:54 * DanC wonders what google thinks "N3" means... timbl's N3 is 2nd on the list ;-)

05:00:33 <AaronSw> .google n3

05:00:34 <xena> n3: http://www.cdex.n3.net

05:02:55 * DanC doesn't see reference to RDF::N3 in that weblog

05:03:07 <DanC> ah.. there it is

05:06:52 <bijan> DanC, I don't know if you caught any of the Topic Map Rantage from yesterday, but i was wondering what you thought about 'em (er..topic maps, not the rantage :))

05:07:19 <DanC> didn't see rantage

05:07:50 <DanC> I scratched my head, hard, about topicmaps in March. Can't say it was a successful effort.

05:07:57 <bijan> It was mostly me :)

05:08:46 <DanC> march trip report: http://www.w3.org/2001/03dc-aus/trip

05:10:06 <mnot> DanC: if I want to write a cwm module, should I be looking at anything in particular besides cwm_string and cwm_os?

05:10:12 <bijan> Interesting.

05:10:35 <AaronSw> I think the best part of that trip was where DanC does his clown dance. ;-)

05:10:58 <DanC> if you want to add built-ins to cwm, that's the place to look, mnot. If you want to teach cwm a different syntax to import/export, I'd point you elsewhere.

05:11:36 <mnot> nope, built-ins is where I'm at. Thx.

05:11:48 * bijan has been dorking with semantic nets and decides he Just Doesn't Really Like Them.

05:12:41 * DanC chalks another one up for The Great CWM Refactoring

05:12:48 <mnot> DanC: re - http://www.w3.org/People/Connolly/drafts/web-research

05:13:04 <mnot> "The cache on each machine should inform the other. Surely there are cache protocols that do this out there; does Jigsaw grok them?

05:13:04 <mnot> "

05:13:17 <mnot> You're wanting cache sync?

05:13:28 <DanC> er.. something like that, yes.

05:13:49 <bijan>http://ep.open.ac.uk/PubSys/resources/html/free0000.html

05:13:49 <dc_rdfig> B: http://ep.open.ac.uk/PubSys/resources/html/free0000.html from bijan

05:13:58 <bijan> B:|Topic Maps vs. RDF

05:13:58 <dc_rdfig> titled item B

05:14:01 <mnot> Closest thing is probably ICP... which is putrid. State of standard cache-related protocols isn't very advanced...

05:14:15 <DanC> e.g. I have wwwoffle running on my desktop and my laptop. If I ask for a page from my desktop, once that wwoffle has got it, I'd like it to contact the laptop and say "here, he'll probably want this while he's on the plane"

05:15:06 <mnot> Hmm, most everything out there now is pull-based.

05:15:50 <mnot> That's actually an interesting use case for RUP in WEBI...

05:16:04 <DanC> RUP: symbol not bound

05:16:05 <dc_rdfig> Label RUP not found.

05:16:24 <mnot> RUP = Resource Update Protocol

05:16:25 * DanC wonders if anybody gets the joke

05:16:47 <bijan> From the paper:

05:16:50 <bijan> RDF provides a model for describing resources. Resources are named using URIs and according to the W3C anything can have a URI (including George Steinbrenner and the Yankees).

05:16:58 <bijan> ARRGH!

05:17:24 * DanC wonders, i.e., if anybody else has spent any time sitting in front of a lisp machine. "symbol not bound" is what you get in response to a typo, ala "command not found" at a unix prompt

05:17:25 <mnot> Is that a URI for the concept of "George Stienbrenner and the Yankees", or one each?

05:17:54 <bijan> Nope, but at an irc chump line noting "Label not found" :)

05:18:30 <mnot> Hmm, there was almost a Haiku up there somewhere...

05:19:39 <DanC> for some reason, "symbol not bound" is the error message that plays in my head whenever somebody uses a buzzword in conversation with me, assuming I know what it means.

05:20:31 <bijan> RUP: symbol not bound; Lisp R-E-P-L returns; (setq (quote RUP) 5)

05:20:31 <dc_rdfig> Label RUP not found.

05:20:42 <DanC> where were we? oh yeah... mnot was going thru my web-research notebook...

05:21:21 <mnot> and musing on the state of caching.

05:22:31 <bijan> B:

05:22:31 <dc_rdfig>http://ep.open.ac.uk/PubSys/resources/html/free0000.html

05:22:32 <dc_rdfig> Topic Maps vs. RDF

05:22:37 <DanC> caching is fun to think about, until you get to deployment, when you realize that web browsers still keep DNS records past expiration, and only pay attention to one A record. Sigh.

05:23:40 <mnot> Yes, that quite sucks.

05:23:41 <bijan> B: I do not like this paper, I do not like it Sam I am. I do not like it Maps Topique. I do not like its RDF-speak.

05:23:42 <dc_rdfig> commented item B

05:24:27 <DanC> LOLOLOLOLOL

05:24:47 * bijan is a little punchy :)

05:25:38 <DanC> ah.. sandro, here's one of my blurbs about naming digital photos: "I use xsane, and I have used sane-cgi, but where do I store the copies, balancing

05:25:38 <DanC> confidentiality, redundancy/availability, and convenience? What URI/name do I give them?

05:25:38 <DanC> How do I associate the RDF transcriptions of them?" -- http://www.w3.org/People/Connolly/drafts/web-research

05:26:48 <DanC> I really want to spend some time playing with freenet. I have a hunch that it's a pretty good answer to a lot of "what do I call the damned thing?" issues.

05:27:27 * bijan is still smarting over the "URIs can name anything...according to the W3C".

05:28:03 <AaronSw> Danc, consensus at p2pcon was that freenet is pretty broken

05:28:05 <DanC> I don't see why you find that so offensive. The attribution is a little iffy; it's actually an IETF spec that says so, but otherwise, what's wrong with it?

05:28:26 <DanC> freenet is broken? bummer. Is there something better?

05:28:39 <AaronSw> mojonation.net is pretty cool

05:28:43 <DanC> my intuition previous to hearing that freenet worked was that it's an impossible problem.

05:29:04 <bijan> Hmm. I find the whole "URIs can name anything, *so* RDF can talk about anything" trope annoying.

05:29:21 <DanC> well, yes, it's annoying.

05:29:21 <AaronSw> why is it annoying?

05:29:39 <bijan> Aaron: because it's either trivial or false.

05:29:57 <DanC> false in the sense that, e.g., not every real number can have a (distinct) URI

05:30:05 <DanC> and trivial otherwise, yes.

05:30:24 <bijan> There's no special representational power that comes with URIs.

05:30:33 <bijan> Any more than, say, with lisp atoms.

05:30:44 <DanC> or integers, for that matter.

05:30:54 <bijan> Indeed.

05:32:54 <AaronSw> hmm, why not the integers?

05:32:55 <bijan> If it were *just* false or trivial, I wouldn't mind. But I think it confuses people.

05:33:28 <bijan> Aaron: you can map URIs onto the integers.

05:33:52 <DanC> who said "not the integers", Aaron? we're saying that the naming power of URIs, lisp atoms, and integers is equivalent.

05:34:15 <bijan> Actually, I should probably say "sexpr" atoms.

05:34:23 <AaronSw> Oh, sorry, mistread your comment

05:34:35 <AaronSw> err misread

05:34:40 <bijan> Since "lisp atoms" might have more representational structure (e.g., CL symbols).

05:35:21 <DanC> the naming power in the web is not intrinsict to URIs; it's in the fact that we got so many agents (humans/machines) to use these silly-looking names in communication with one another, and thereby bound them to useful concepts.

05:35:40 <DanC> ooh. was that profound?

05:36:05 <bijan> Anyhoo, I've now idenitfied the *Real Reason* the appearence of that trope in this paper torqued me off...I was hoping for a more formal and precise comparasion between the two, whereas the paper is all fuzzy.

05:36:54 <DanC> indeed, I've been disappointed at the level of rigor in topicmaps stuff many times.

05:37:17 <DanC> I get the same "where's the beef" feel about UML.

05:37:44 * bijan notes that this happens with RDF stuff too...although good progress has been made, especially in the last 8 months, IMHO.

05:38:35 <bijan> It's good to see the RDF community moving *away* from the Topic Map type fuzziness.

05:38:39 <DanC> yes, I'm a little embarrased at the (low) level of rigor in some earlier RDF stuff, but I'm quite happy with recent progress.

05:39:18 <DanC> embarrassed because at the time, I thought the RDF stuff was quite rigorous. then I got a few clues

05:40:31 <DanC> moving away from fuzziness appeals to me, but it's only an unmitigated good thing if we're also getting more tools and content deployed.

05:40:39 <bijan> Rigor, (various sorts of) clarity, and understandability don't always march together :)

05:40:42 * sandro (not quite asleep) is very happily surprised that the "informal" syntax that the WSDL 1.1 spec uses to define WSDL matches almost exactly the syntax he (err, I) way playing with the blindfold grammars for XML.

05:41:02 <DanC> interesting, sandro.

05:41:38 <sandro> yeah. great minds, or something. :-)

05:42:06 <sandro> I'll have to see if I get my code to read their spec cut & pasted or something.

05:42:17 * DanC suspects I've descended into navel-gazing and should knock off for the night

05:42:41 * AaronSw curses how domainmonger has broken the back button -- their site used to be so good

05:42:44 <sandro> ditto; gnight.

05:42:48 <DanC> the premise of tonight's session was PerfectOffice research... choosing my desktop platform for the next few years.

05:44:21 <DanC> DanC has changed the topic to: Life, the Semantic Web, and Everything (http://rdfig.xmlhack.com/)

05:45:04 <bijan>http://xml.coverpages.org/RDF-TopicMaps-LateLazyVersusEarlyPreemptiveReification.html

05:45:04 <dc_rdfig> C: http://xml.coverpages.org/RDF-TopicMaps-LateLazyVersusEarlyPreemptiveReification.html from bijan

05:45:24 <bijan> C:|Another RDF/Topic Map comparison...

05:45:24 <dc_rdfig> titled item C

05:45:31 <bijan> C:...that doens't do me a damn bit of good.

05:45:31 <dc_rdfig> commented item C

05:45:36 * DanC wonders what flew into bijan's bonnet and got him wandering thru topicmap-land

05:45:44 <AaronSw> bijan, have you seen larsbot's paper?

05:45:50 <bijan> Check the rantage from yesterday :)

05:45:53 <bijan> aaron: no.

05:46:15 <DanC> phpht. no link from http://www.w3.org/RDF/Interest/#irc to the logs.

05:46:36 <bijan> Basically, someone said that what will distinguish Topic maps from RDF is the apps.

05:46:42 <DanC> the /RDF/Interest/ page shouldn't give the impression that it's frequently maintained. (i.e. shouldn't have a "News!" heading)

05:46:56 <bijan> Something like the "apps they each are suited for".

05:47:03 <bijan> That just seemed like baloney to me :)

05:47:11 <AaronSw> hmmph, don't have the link on my machine

05:47:26 <AaronSw> err this machine

05:48:03 * DanC can't find "topicmap" in yesterday's log

05:48:41 <bijan> Hmm. I thought it was yesterday...

05:49:06 <DanC> ah... "topic map". note the space. http://ilrt.org/discovery/chatlogs/rdfig/2001-11-19.html#T22-28-59

05:50:37 <DanC> best RDF app to point to, btw, is PRISM, I think. and RSS, I guess. Dublin Core, sorta. Oh! and DMOZ, sorta. (cf http://www.w3.org/RDF/#projects)

05:51:01 <AaronSw> PRISM?

05:51:18 <DanC> PRISM... yeah... the thing Ron Daniel is working on.

05:51:39 <DanC> look under http://www.w3.org/RDF/#agg

05:51:45 <AaronSw> I know what it is... I just never thought of it as an exemplary RDF app

05:52:12 <AaronSw> Hmm, why would you point to it, and not, say, MusicBrainz?

05:52:18 <DanC> no? (a) it's commercially relevant [if you believe what they say] (b) it actually mixes vocabularies

05:52:26 <AaronSw> Hmm.

05:52:41 <DanC> MusicBrainz is good too

05:53:10 <AaronSw> I saw Laird Popkin trying to sell PRISM to the "Napserize the Enterprise" guys at P2PCon

05:55:02 <AaronSw> Well, I got my backup started again, so I think I'm going to go to sleep.

05:59:26 <AaronSw> nite all

05:59:30 <bijan> Nite.

06:00:01 <AaronSw> c'ya

06:00:02 * AaronSw disconnects

08:27:33 <mkl> mkl is now known as mario

09:04:36 <lilo> [GlobalNotice] Apologies, all. Apparently one of the rotation servers was hubbed to the one we restarted.

10:22:59 <ddent> [GlobalNotice] Hiya, one little rehub now and then thats it for the evening. Sorry.

10:50:38 <danbri> dc_rdfig:view

10:50:39 <dc_rdfig> A: http://www.isacat.net/svg/showme.html

10:50:40 <dc_rdfig> B: Topic Maps vs. RDF (http://ep.open.ac.uk/PubSys/resources/html/free0000.html)

10:50:41 <dc_rdfig> C: Another RDF/Topic Map comparison... (http://xml.coverpages.org/RDF-TopicMaps-LateLazyVersusEarlyPreemptiveReification.html)

10:56:47 <danbri> B:A really dissapointing paper. Full of silly posturing, pretty unhelpful all round. "the significantly greater knowledge-modeling power of ISO topic maps requires more thought on the part of its users", indeed. CycL and Unicode, in their own way, have greater knowledge modeling power than RDF too...

10:56:47 <dc_rdfig> commented item B

10:58:35 <larsbot> danbri, I'm hoping to be able to rewrite my own RDF vs TM paper soon

10:58:42 <larsbot> do you have any comments on it?

10:58:51 <danbri> I look foreward to it :)

10:59:15 <danbri> comments? not yet, no... (except from what I saw it was a lot more constructive and technically interesting than B: above...)

10:59:52 <larsbot> it's written from a topic maps perspective (unavoidably)

11:00:00 <larsbot> so comments from RDF people would help me get it right

11:00:13 <larsbot> in short: comments much appreciated, especially before the rewrite :-)

11:00:42 <danbri> I'll try to find time to comment, but am pretty snowed under... You might try a draft out on www-rdf-interest, there's a lot more people there than in #rdfig...

11:01:03 <larsbot> good point

12:25:11 <danbri>http://www.w3.org/History/1991-WWW-NeXT/Implementation/

12:25:12 <dc_rdfig> D: http://www.w3.org/History/1991-WWW-NeXT/Implementation/ from danbri

12:25:44 <danbri> D:|TimBL's original NeXT WorldWideWeb browser, source code

12:25:44 <dc_rdfig> titled item D

13:21:06 <tim-gone> D: Some of the .h files still dated Oct 1990 ... surprised they haven't been touched since.

13:21:07 <dc_rdfig> commented item D

13:24:18 <danbri> D:We need a .nib convertor from the old NeXT format to something GNUStep or MacOS X will understand...

13:24:19 <dc_rdfig> commented item D

13:26:17 <larsbot> bijan, are you there?

13:28:37 <danbri> D:From the [bug list|http://www.w3.org/History/1991-WWW-NeXT/Implementation/Bugs.html] "Not included yet but won't be missed yet: Link Typing"

13:28:37 <dc_rdfig> commented item D

13:32:36 <larsbot> this source is objective C, right?

13:32:49 <tim-gone> Bulding a new interface would be a piece of cake, if the NeXTStep interface is the same for the software mosules.

13:33:17 <tim-gone> tim-gone is now known as tim

13:36:19 <danbri> I don't think they've changed things much... though things are called NSFoo instead of NXFoo... Damian was taking a look, said we'd need an understanding of the contents of the .nib (a format that they never documented) to know what needed doing

13:46:19 <jhendler-h> B: Basically, all that are needed are standard ontologies, probably expressed on the Web as topic maps.

13:46:19 <dc_rdfig> commented item B

13:46:38 <jhendler-h> B: Or maybe those ontologies will be expressed in RDF!

13:46:39 <dc_rdfig> commented item B

13:57:53 <jhendler-h> (what's a "semi-laugh" icon)

13:58:08 <jhendler-h> i.e. the sort of something is said ironically...

14:03:24 <larsbot> <irony>...</irony>, perhaps :)

15:15:31 <timbl> :-/

15:57:44 <danbri> hey gerald...

15:57:55 * danbri gerald and sbp move a convesation over from #validator

15:57:59 * sbp waves... what's going on over here?

15:58:25 <danbri> not much :)

15:58:57 <danbri> So, a sha1 property... should we / could we define it as daml:UnambiguousProperty? (since sha1 isn't really, but astronomical chance of collison)

15:59:55 * danbri gerald and sbp wander off in various directions

16:00:23 <gerald> danbri, care to contrast http://lists.w3.org/Archives/Public/www-archive/2001Nov/0045.html with p2p stuff, e.g. http://impressive.net/archives/fogo/20011025140605.B20764@impressive.net

16:00:59 <gerald> e.g. how could/should p2p systems use sha1 properties to identify resources, vs some kind of sha-1 URI scheme

16:01:23 <sbp> well, is it a good enough approximation is the question...

16:01:44 <danbri> ah, the folks on #infoanarchy mentioned that when I went there to ask the p2p hackers what conventions they use for writing sha1s

16:02:01 <danbri> apparently the CAW guy doesn't like http extension framework, which is a pity

16:02:30 <sbp> if you declare it as Unambiguous, then any software that uses the property will have to put up with the hash collosion problem, which isn't really a bad thing... using SlightlyLessThanUnambiguous simply means that the processor has to note that soemtimes there might be a collision

16:03:43 * danbri re-reads http://impressive.net/archives/fogo/20011025140605.B20764@impressive.net

16:03:59 <danbri> good, gerald said everything that i had in mind to say except never got around to :)

16:04:34 <gerald> but I wrote "I think URIs based on sha-1 hashes are a fantastic way to identify

16:04:34 <gerald> resources in P2P systems" and it doesn't seem you agree

16:06:09 <danbri> Well, I think identifying resources using sha-1 is fantastic; and if anyone has the patience to arrange the URI scheme that's good. I'm lazy: identifying them with a pair of (the sha1 property uri;some sha1 value) is good too

16:07:00 <gerald> ok, I see.

16:07:15 <sbp> you could even say: { ?x util:sha1 ?y } log:implies { ?x log:uri [ log:concat ("hash:sha-1:" ?y) ] } .

16:07:51 <sbp> then have the best of both worlds :-)

16:08:19 * gerald wanders off to make coffee

16:09:54 <sbp> of course, that would mean that the hash scheme denotes the thing that *has* that hash value, and not the hash itself. Is that what we want?

16:10:26 <bijan> What happens if you have multiple @prefixs declaring the same prefix in the same document?

16:11:34 <sbp> Er... TimBL documented that somewhere... let me check...

16:11:36 <danbri> according to the Notation3 Language Specification (aka Cwm? ;-), try it and see...

16:12:25 <sbp> "Once set up, a prefix can be used for the rest of the file (* may change)." - http://www.w3.org/2000/10/swap/Primer

16:12:30 <sbp> according to the documentation...

16:13:32 * bijan pegging a little progress on prolog/n3

16:13:33 <bijan> Er...

16:13:46 <bijan> "can be" and *what* may change, the prefix or that it's used for the rest of the file?

16:15:32 <sbp> ah, on the contrary: CWM lets you change it

16:15:51 <sbp> I took TimBL's statement to mean that a prefix is fixed for the rest of the document

16:15:59 <sbp> CWM 1.80 doesn't seem to agree

16:16:48 <sbp> In general, it's best to follow CWM, because it's more up-to-date than the documentation

16:16:51 <sbp> Gotta run

16:19:30 <larsbot> bijan, I've been thinking a bit more about the term "reification"

16:19:35 <larsbot> you ready to talk a little about it?

16:19:51 <bijan> very little.

16:19:58 <larsbot> may be all we need

16:20:10 <bijan> I mena, what's to talk about? Stupid bit of jargon ;)

16:20:10 <larsbot> is reification a special kind of representation?

16:20:14 <bijan> No.

16:20:36 <bijan> not even reification(tm)

16:20:47 <larsbot> so if A reifies B then A does not in any way, shape, or form also represent B?

16:20:56 <bijan> Huh?

16:21:05 <larsbot> if A reifies B, does it then represent B?

16:21:15 <bijan> Look, reification(tm) is a perfectly ordinary form of representation

16:21:27 <bijan> That's why I think reificaiton(tm) should be junked as a term.

16:21:42 <larsbot> never mind that; I want to learn what the term means

16:21:53 <bijan> *Which term*?

16:21:54 <bijan> There are several.

16:21:57 <bijan> refification(tm)

16:21:59 <larsbot> reification as used in AI

16:22:07 <bijan> Ah.

16:22:26 <bijan> Well, I only know it in terms of RDF and Loom.

16:22:49 <bijan> Take rdf.

16:22:55 <larsbot> I have an RDF triple, and create a resource to represent it

16:22:56 <danbri> please!

16:23:06 <larsbot> that resource reifies the triple, right?

16:23:10 <larsbot> (please what?)

16:23:27 <bijan> Yes.

16:23:27 <danbri> (please)"Take rdf." :)

16:23:33 <larsbot> (ok :)

16:23:44 <bijan> Er...actually it's a little trickier.

16:23:48 <larsbot> but it also *represents* the triple, does it not?

16:23:51 <danbri> that resource _represents_ the triple; the term 'reification' adds little but confusion.

16:24:11 <bijan> Actually, I'd say it *expresses* the triple :)

16:24:12 <larsbot> yeah -- I'm trying to remove that confusion from within my head

16:24:34 <bijan> look, I could dub any ole statement http://thatdamnstatemetn

16:24:49 <danbri> There's also an active sense of the word; less nouny, where we talk about reification as something parsers etc do, the 'make a reification/representation of' some stuff.

16:25:02 <bijan> but that would be outside rdf

16:25:15 <danbri> So part of the confusion comes from the loose use of langauge, where the representation and the act of representing, both get termed 'reification'

16:25:19 <bijan> In the sense that I couldn't use *that* representation in regular inferences as well.

16:25:49 <larsbot> can resource A only represent B if B is part of an RDF model?

16:25:50 <bijan> It's an interesting question whether a reified statement is also a *quoted* statement.

16:26:35 <larsbot> that is, if B is something outside an RDF model, say a file or a squash racket, is it then impossible for some RDF resource to reify it?

16:26:51 <bijan> reify(rdf) yes.

16:27:01 <bijan> you can only reify statements.

16:27:07 <bijan> Indeed, only reify rdf statements.

16:27:39 <bijan> This paper discusses the semantics and usage of reification as applied to relations and tuples. The reification of a tuple is a proposition object possessing a case role for each domain attribute in the tuple. The reification of a set of fillers of a role

16:28:00 <bijan> is an abject sometimes referred to as a "roleset".

16:28:45 <larsbot> uh, what paper?

16:29:02 <bijan> Loom paper: http://www.isi.edu/isd/LOOM/papers/macgregor/propositions.html

16:29:15 <larsbot> thanks!

16:29:38 <bijan> Searching for "reification loom" turns up some useful stuff, imho.

16:29:52 <larsbot> what's the relationship between an RDF resource (say, tag:...) intended to denote me, and me?

16:30:16 <bijan> The URI names you.

16:30:21 <bijan> I think you *are* the RDF resource :)

16:30:40 <danbri> (from the implicit point of view of some piece of rdf)

16:30:42 <larsbot> well, there's something inside the RDF model that denotes me, and it certainly isn't me!

16:30:46 <nephrael> nephrael is now known as dmiles

16:30:53 <bijan> The URI.

16:31:01 * sandro agrees he is the RDF resource. "resource" is the RDF term for "object in the domain of discourse"

16:31:08 <larsbot> ah, I get it

16:31:25 <larsbot> the URIs name the resources, right?

16:31:31 <bijan> yes.

16:31:57 <sandro> Yep. (strictly they are URI-References, not URIs -- they usually have fragment IDs.)

16:32:18 <larsbot> so when I create a URI to reify an RDF statement, that URI also names the statement, right?

16:32:29 <bijan> yes.

16:32:41 <sandro> That is: they are a pair of a string that might get you web content AND a name of something perhaps within that content.

16:32:45 <larsbot> so reification is a specialization of naming, not of representation?

16:32:56 <bijan> Naming is a kind of representation.

16:33:07 <larsbot> that makes good sense

16:33:15 <larsbot> and of course, topics do *not* name their subjects

16:33:22 <bijan> no, they do.

16:33:26 <bijan> They *are* the name of thier subjects.

16:33:33 <larsbot> are you joking?

16:33:35 <bijan> No.

16:33:46 <bijan> Topic map folks are confused.

16:34:17 <larsbot> if I have a topic, and it has a subject indicator, and the resource used as a subject indicator describes me, then how does the topic name me?

16:34:33 <sandro> Bijan, can you stipulate that in a "Topic maps" conversation, the default namespace is of terms as-defined-by the Topic Maps spec?

16:35:06 <bijan> sandro: I could, but it makes using words like "name" difficult.

16:35:23 <bijan> And larsbot would have too, too.

16:35:26 <larsbot> sandro, I would say so, but it would be better if the terms in that namespace were as like their namesakes outside that namespace

16:35:40 <bijan> Since he's using "name" equivically.

16:35:40 <larsbot> as possible

16:36:16 <bijan> larsbot, re: your question, the topic name refers to you.

16:36:19 <bijan> Er..

16:36:23 <bijan> The topic refers to you.

16:36:32 <larsbot> yes, it does

16:36:39 <bijan> the reference (naming) relation is via the subject indicator.

16:36:51 <larsbot> referencing and naming is not the same

16:37:00 <larsbot> or is it?

16:37:07 <larsbot> no, it's not

16:37:09 <bijan> A name refers to that which it names.

16:37:18 <bijan> A name is a referring expression.

16:37:20 <larsbot> yeah, but one can refer to things without naming them

16:37:27 <larsbot> so naming is a specialization of referencing

16:37:29 <bijan> Sure.

16:37:36 <bijan> To the form.

16:37:47 <bijan> Topics sure look like names.

16:38:01 <larsbot> a topic is just a point, a node

16:38:19 <bijan> They are idenitifers, singular terms, and, from what I can tell, they refer via a baptismal event.

16:38:27 <larsbot> it has characteristics attached to it, but the node itself is without properties

16:38:32 <bijan> <sigh/>It has identity.

16:38:34 <bijan> It refers.

16:38:40 <larsbot> it does have identity, yes

16:38:52 <larsbot> what is the relationship between referring and representing?

16:38:52 <bijan> Right, so that identity stands for the subject.

16:38:59 <larsbot> right

16:39:08 <bijan> Thus it names the subject, this isn't hard ;)

16:39:21 <bijan> names(NTM)

16:39:27 <larsbot> but the topic has no syntactical existence

16:39:47 <larsbot> it's either an abstraction or an object in memory or an RDBMS row

16:39:47 <bijan> I don't know what that means.

16:40:15 <larsbot> what I mean is that the name is attached to the topic, but is not part of it

16:40:21 <larsbot> so it's the name that names the subject, and not the topic

16:40:38 <bijan> A topic is a resource that acts as a proxy for some subject; it is the topic map system's representation of that subject.

16:40:46 <larsbot> yes

16:40:47 <bijan> No, name(tm) is just a label of the topic.

16:41:04 <larsbot> a label of the subject, which the topic represents

16:41:36 <bijan> In order to discourse on a subject within the topic map paradigm, that subject must be reified through the creation of a topic.

16:41:40 <bijan> (from the spec)

16:41:45 <larsbot> yes, that's right

16:41:47 <bijan> In order to *discource*.

16:41:51 <larsbot> yes

16:41:52 <bijan> I.e., to *talk about*.

16:42:00 <larsbot> where are you headed, bijan?

16:42:02 <bijan> Ergo, a topic is a representational device.

16:42:08 <bijan> In particualr, it's a singular term.

16:42:15 <larsbot> yes, I've already said that

16:42:18 <larsbot> and yes, it is singular

16:42:32 <bijan> It's a singular term, afaict, without structure (i.e., it's not a definition description)

16:42:36 <bijan> Ergo, it's a name.

16:42:53 <bijan> Indeed, technically, it's a "constant" or "constant name".

16:43:05 <larsbot> name in a wider sense of the term, then

16:43:13 <bijan> name(NTM)

16:43:37 <larsbot> but what *is* the relationship between referencing and representing?

16:43:47 <larsbot> that's the key issue, I think

16:43:50 <bijan> no.

16:43:55 <bijan> It's not "referencing" it's reference.

16:44:22 <larsbot> what's the "it" here?

16:44:50 <bijan> "referencing" sounds like looking something up in a dictionary.

16:45:08 <bijan> The relation is reference.

16:45:30 <larsbot> what's the relationship between to reference and to represent, then?

16:45:42 <bijan> Complicated.

16:45:51 <larsbot> no surprise there :-)

16:46:26 <larsbot> let's say that we have a topic, and its subject is an association in a topic map

16:46:28 <bijan> By and large, vastly simplifying, the reference relation is the "aboutness" relation between a representation and what it represents.

16:46:46 <larsbot> what's the relationship between the topic and its subject in this case

16:46:51 <larsbot> bijan, that makes perfect sense

16:47:29 <bijan> Following, frege, we can distinguish at least one other componant of (linguistic) representation, "sense".

16:47:51 <bijan> Where "sense" roughly is the "meaning" of the expression.

16:47:51 <larsbot> "sense", as in "meaning"?

16:47:53 <tim> catching ip ...sha1 uri scheme good idea. Multiple @prefixes for same namespace is fine, cwm picks random one on output. Multiple @prefixes for same prefix, diff namespace best avoid. May change: the rest of the file. Could forbid clashing later, or could even start using scoping to alloweasier embedding of self-contained formulae.

16:48:39 <bijan> tim, so it's kosher to reject a document that reassigns a prefix?

16:48:52 <larsbot> btw, there's a "#topicmaps" channel on this server, dedicated to topic maps

16:49:06 <larsbot> we could take this discussion there, as it seems more TM-related than RDF-related right now

16:49:39 * bijan notes that he doesn't really *want* to discuss TMs :)

16:50:06 <larsbot> fair enough :)

16:50:16 <larsbot> let's drop it, then

16:50:46 <bijan> I will say this, the interesting bit I've not seen is a explanation of the actual expressive power of TMs.

16:50:54 <bijan> I.e., something akin to the RDF model theory.

16:51:15 <danbri> the TM query folk had some mathsy papers floating around, I vaguely recall

16:51:57 <tim> Lars, "well, there's something inside the RDF model that denotes me, and it certainly isn't me!" Inside the "model"? In general, I think there is the syntax and there is you. That's it.

16:52:01 * tim caught up

16:52:30 <larsbot> tim, I was talking about one specific RDF model instance that had statements about me in it

16:52:42 <larsbot> what you get when you parse foo.rdf is an RDF model, isn't it?

16:53:04 <tim> bijan, so it's kosher to reject a document that reassigns a prefix? I would be tolerant -- maybe issue a warning. There is a nice property that a concatenation of two valid N3 documents is a valid N3 document.

16:53:07 <sandro> Calling it a RDF Graph is probably less confusing.

16:53:19 <larsbot> I'll do that in the future, then :)

16:53:28 <sandro> "Model" means many things in very-slightly-different contexts.

16:53:45 <larsbot> bijan, I might like to see that, too, once I understand what model theory does :-)

16:53:53 <larsbot> I'm not going to plague this channel with that, though

16:54:08 <bijan> It specifies the semantics of the language in formal (i.e., mathematical) terms.

16:54:12 <larsbot> danbri, I guess you're referring to one of Ann Wrightson's early papers

16:54:21 <larsbot> bijan, I know that, but no more

16:54:23 <tim> There is an abstract thing, and RDF graph. You are a node in it. You don't really "get" it from a parser. What you get from a parser depends on what your parser parses to.

16:54:40 <larsbot> true, but the parser creates a representation of something

16:54:44 <larsbot> just like an XML parser does

16:54:45 <danbri> larsbot, yes, I think it was Ann's

16:54:54 <tim> Suppose we call the thing in a language which represnents an RDF graph a formula.

16:56:06 <tim> The parser does craete something -- like a structure in C or something. That isn't really addressed by the RDF spec. It would be if we had an RDF API -- RDF DOM if you like.

16:56:18 <bijan> NOOOOOOOOO!

16:56:21 <bijan> :)

16:56:35 <larsbot> tim, I get this; it was the term "rdf graph" I was looking for

16:56:53 <larsbot> so, in the rdf graph, there was a node, and that node represented me

16:57:00 <tim> No, that node is you.

16:57:05 <bijan> Er...

16:57:15 <larsbot> so, what's the thing in the graph called, then?

16:57:27 <danbri> from RDF's point of view. The RDF file may be wrong... or a lie...

16:57:31 * danbri not helping

16:57:37 * larsbot agrees :)

16:57:38 <bijan> I think larsbot is taking RDF Graph to mean the in memory structure.

16:57:39 <sandro> Tim, I think the node represents him.

16:57:57 <tim> (I am happy to have two levels linked by a 1:1 isomorphism or one level here, but I don't normally like to add a level except fro good reason)

16:58:02 * larsbot thinks he takes the RDF Graph to be the thing represented by the in-memory structure

16:58:20 <bijan> Ah, then it's likely that you are the node.

16:58:36 <tim> Ok, I'll go along with larsbot:represents as a mapping between the nodes in an abstract rdf graph and the things which are actually described by the RDF.

16:59:17 <larsbot> bijan, in one sense yes, and in another no

16:59:26 <larsbot> tim, good :)

16:59:43 <bijan> Well, I did say "likely" :)

16:59:51 <larsbot> I noted that :)

16:59:57 <bijan> Assumign a standard interpretation.

17:00:26 <bijan>http://www.semanticweb.org/SWWS/program/full/paper53.pdf

17:00:27 <dc_rdfig> E: http://www.semanticweb.org/SWWS/program/full/paper53.pdf from bijan

17:00:58 <tim> So - meaning: Joe loves Mary. Graph: There is arc representing love from node represneting Joe to node represneting Mary. Syntax: ":joe :loves "Mary."

17:01:10 <bijan> E:|Somewhat better (although PDF) paper on Topic Maps and RDF

17:01:11 <dc_rdfig> titled item E

17:01:41 <larsbot> bijan, that's the Decker and Lacher paper, right?

17:01:45 <bijan> Yes.

17:01:48 <larsbot> tim, I agree completely

17:01:51 <danbri> tell it to the chump!

17:02:02 <larsbot> what, me?

17:02:10 <danbri> E:by Decker and Lacher

17:02:11 <dc_rdfig> commented item E

17:02:12 <danbri> anyone

17:02:14 <danbri> :)

17:02:15 <larsbot> ok :)

17:03:14 * danbri tries to interest timbl in getting http://www.w3.org/History/1991-WWW-NeXT/Implementation/ running under MacOS X...

17:03:35 <bijan> RDF requires additional vocabulary such as DAML+OIL for

17:03:35 <bijan> ontology definition and RDF itself merely provides the object layer in this data model stack.

17:03:36 <danbri> retro-chic...

17:03:37 <bijan> Topic Maps on the other side have richer semantics, and provide a number of features of an

17:03:46 <bijan> ontology definition language. For a comparison on the semantic layer, DAML+OIL based on

17:03:48 <bijan> RDF is a more appropriate candidate for a comparison with Topic Maps. However, this will

17:03:48 <bijan> not be investigated in this paper.

17:04:35 <bijan> I don't think I actually mentioned that when I was ranting about them the other day. Mostly because I'm still not sure what "level" Topic Maps want to be at :)

17:05:54 * tim already has a NeXTStep box at work and a linux box at home which he hasn't had the time to keep up to speed

17:06:22 <danbri> bummer

17:06:47 <tim> Could add OsX to the list ...

17:06:57 <danbri> was the WorldWideWeb code opensourced, like Cern server...? can people just go play with it...

17:07:03 <bijan> danbri: why not point to that on a nextdev/osX dev list.

17:07:23 <bijan> Get some OsX/nextstep jocks working on it for the sake of bragging rights?

17:07:42 * bijan 's motto: Never do yourself what you can get someone else to do for you.

17:07:49 <larsbot> bijan, you may want to look at: http://www.garshol.priv.no/tmp/tmrdf-paper.html

17:08:06 <danbri> shellac(damian) already asked on the gnustep list about tools for porting... but didn't point as src as we weren't sure if it was properly opensource

17:08:17 <larsbot> that paper is going to be rewritten (hopefully this week), so you can even get me to correct the mistakes you find in it :)

17:08:17 <danbri> will do though

17:08:27 <larsbot> please do *not* chump it yet, though

17:08:36 * sandro wonders how bijan's motto applies to pleasurable experiences (like programming :-)

17:11:03 <bijan> Well, it applies perfectly well to pleasurable experiences....*like programming*.

17:11:26 <bijan> Pleasurable experiences like, oh, *eating* are a slightly different matter :)

17:11:33 <tim> WorldWideWeb is open source.

17:11:38 <bijan> larsbot, Examples of such things may be the person 'Lars Marius Garshol', the company 'Ontopia', and this paper.

17:11:47 <larsbot> yes?

17:11:53 <danbri> "<tim> WorldWideWeb is open source." :-)

17:12:09 <bijan> surely you mean the person, Lars Marius Garshal, the company, Ontopia,

17:12:17 <danbri> let's see if any of the next/osx folk want to play with it then...

17:12:25 <bijan> Or "the person named 'Lars...'"

17:12:36 <larsbot> bijan, quite possibly

17:13:30 <bijan> I dont' see why you separate DAML and OIL, since I only really know aobut DAML+OIL :)

17:13:53 <larsbot> bijan, it seems that I was using obsolete information

17:13:54 <danbri> that's what I said. Its a recent history thing; start of this year they were more distinct.

17:14:04 <danbri> been a busy year.

17:14:07 <tim> Surely, you mean "surely you mean '...the person, Lars Marius Garshal, the company, Ontopia,...' "?

17:14:14 <bijan> yes.

17:14:25 <larsbot> daml+oil: that's one of the things I intend to fix

17:14:28 <bijan> The quotes were left out as an exercise for the reader.

17:14:35 <larsbot> another is the statement that Ontopia's home page employs my mailbox :)

17:14:52 <bijan> In RDF statements, in the form of (subject, property, object) triples, are the only way of assigning characteristics, while in topic maps topics may have names, occurrences, and participate in associations.

17:15:01 <bijan> That seems silly.

17:15:32 <larsbot> you are perfectly free to think so

17:15:33 <bijan> That doesn't seem like a real distinction to me.

17:15:45 <larsbot> in software it is very real

17:15:57 <bijan> the stuff topic maps do is what would be done in a rdf vocabulary.

17:16:03 <larsbot> not all of it

17:16:05 * danbri realises why he hates Xchat; it forces me to click to see the new stuff in different irc channels; which encourages useless flailing and distraction from task at hand.

17:16:22 * larsbot agrees

17:16:49 * danbri makes a break for productivity; bye all

17:17:45 <larsbot> there are several distinctions, but I don't present them well enough in the paper

17:17:49 <larsbot> one relates to merging

17:17:53 <larsbot> another to subject identity

17:18:22 <larsbot> a third to how the form of the model makes you do things differently from how you would in RDF

17:18:26 <bijan> sure, there's lots of things that RDF doesn't define. I agree it's lower level.

17:18:32 <larsbot> of course, at some level they are equivalent, but so are *all* data models

17:18:34 <tim> My cuurent assumption is that the "extra richness" of topic maps can indeed be expressed in RDF, and that the simplicity of RDF makes it the only candidate for the basic architecture. Are there things in TM which break that or map to RDF really clumsily?

17:18:49 <larsbot> tim, I'm still trying to explore that

17:18:56 <larsbot> I don't know RDF well enough to answer completely

17:19:02 <bijan> But being able tot type your properties make the "restriction" of RDF as compared to TM otiose.

17:19:23 <larsbot> what do you mean by "type your properties"?

17:19:51 <bijan> Indeed, I'd say *TMs* are more restrictive as they settle the ways of "assigning characteristcs"

17:20:00 <bijan> In RDF statements, in the form of (subject, property, object) triples, are the only way of assigning characteristics,

17:20:03 <bijan> take that part.

17:20:18 <larsbot> bijan, yes, they are in a sense more restrictive

17:20:20 <larsbot> that's their power

17:20:34 <bijan> What exactly is the "onliness" of that "way of assigning characteristics"?

17:20:42 <bijan> while in topic maps topics may have names, occurrences, and participate in associations.

17:20:56 <bijan> What exactly is the expressive gain? AFAICT, there isn't any.

17:21:06 <bijan> The Decker paper shows a straightforward mapping.

17:21:07 <larsbot> there is no expressive gain that I know of

17:21:16 <larsbot> yes, but that's not interesting

17:21:23 <bijan> Why not?

17:21:28 <larsbot> you could do with RDBMSs, XML, CSVs or anything else you like

17:21:38 <larsbot> *all* data models are equivalent

17:21:46 <larsbot> in terms of expressive power

17:21:56 <bijan> No.

17:22:10 <larsbot> ok, what can you do with RDF that you can't do with an RDBMS?

17:22:13 <larsbot> name a single thing

17:22:42 <tim> Given that, systems which try to lead you to work or organize your system in specific ways are a pain. Not minimalistic. Not a good foundation.

17:22:55 <larsbot> not if they lead you in the right direction

17:23:01 <bijan> That's not the kind of expressiveness I'm talking about.

17:23:04 <tim> (that=all being equiv)

17:23:35 <larsbot> tim, it's the same reason you choose SQL for some things and C for others

17:23:41 <larsbot> bijan, I know

17:23:53 <tim> all data models are equivalent if by data models you mean relational systems.

17:24:09 <bijan> Hmm.

17:24:16 <bijan> no.

17:24:31 <tim> RDF and RDB are equivalent - RDF is just a webized version. N3 rules without nested formulae are SQL-equivalent basically.

17:24:43 <bijan> Datalog is strictly less expressive than Prolog.

17:25:15 <larsbot> they aren't data models; they are programming languages

17:25:38 <bijan> There's a difference? They're declarative programming languages.

17:25:45 <larsbot> yes, but not data models

17:25:52 <larsbot> RDF is a data model

17:25:55 <larsbot> DAML+OIL is not

17:26:00 <larsbot> topic maps are a data model

17:26:02 <larsbot> RDBMSs to

17:26:04 <larsbot> o

17:26:06 <larsbot> and XML

17:26:22 <larsbot> (well, strictly speaking XML has one, but isn't one)

17:26:42 <bijan> Provide a definition of "data model" so I understand the "strictly speaking".

17:26:58 <bijan> Like lisp, prolog has uniformity between code and data.

17:27:10 <larsbot> a data model is a set of data building blocks, and rules for how you can combine them

17:27:29 <bijan> Sounds like Datalog vs. Prolog to me.

17:27:34 <larsbot> lisp has a data model inside it, but it isn't one

17:27:51 <larsbot> the car cdr model (with all the extensions of most lisps) is a data model

17:27:56 <larsbot> the language sits on top of that

17:28:11 <bijan> The difference between Datalog and Prolog isn't the inference model (per se) but the structure of acceptible (i.e., legal) clauses.

17:28:21 <bijan> I.e., the data model.

17:28:25 <larsbot> explain

17:28:36 <bijan> To be more precise, the main difference between datalog and prolog is in their data model :)

17:29:11 <larsbot> what *os

17:29:15 <bijan> Prolog allows structures (functions) as arguments in the head of a clause. I.e., it has unifiable records.

17:29:52 <larsbot> that might be a data model difference; I don't understand Prolog well enough to tell

17:30:21 <larsbot> that it really limits expressive power *of the data model* and not *of the data model + programming language combination* I am not sure

17:30:23 <bijan> What I'm trying to say is there isn't a huge data model/programmign language distinction, espeically when dealing with symbolic processing.

17:30:47 <larsbot> it's true that it's much smaller for Prolog than for Java

17:31:43 <larsbot> you could almost certainly create your own convention for emulating the missing data model capability inside datalog

17:31:49 <bijan> No.

17:32:02 <bijan> That's why prolog is strictly more expressive.

17:32:16 <larsbot> why not? you could have your own set of predicates that allowed you to represent prolog clauses inside datalog

17:32:37 <larsbot> it probably would be horrible to use, but you could in principle

17:32:51 <bijan> No, that's what I'm saying. You couldn't in principle.

17:33:06 <bijan> That's the point.

17:33:14 <larsbot> as long as you have predicates you could

17:33:31 <bijan> you need a proof of that, and you won't get one :)

17:33:38 <bijan> Datalog isn't turing complete, prolog is.

17:33:48 <larsbot> if datalog isn't turing complete, that explains it

17:34:01 <larsbot> could datalog *data* represent RDF?

17:34:07 <bijan> Saying that it's strictly less expressive "explained" it too.

17:34:27 <larsbot> no, it just made the statement, with no explanation

17:34:35 <larsbot> but could it?

17:34:51 <bijan> no, saying that it's not turing complete is just a statment too :)

17:34:52 <bijan> Sure.

17:35:04 <larsbot> well, you can represent prolog programs as RDF

17:35:15 <larsbot> so you could then represent them as datalog structures too

17:35:21 <bijan> <sigh/>

17:35:35 <larsbot> well?

17:36:20 <bijan> You need to read some of the back archives of rdf-logic (i think).

17:36:30 <larsbot> was what I said right, or was it wrong?

17:36:41 <bijan> I can "represent" prolog programs as URIs.

17:36:47 <larsbot> yes

17:36:48 <bijan> But it's not a *useful* representation.

17:36:51 <larsbot> no

17:37:03 <bijan> I can represent Prolog programs as integers.

17:37:07 <larsbot> yes

17:37:13 <bijan> Again, those represtations aren't *equally expressive*.

17:37:23 <bijan> So, yes you can represent prolog programs in RDF.

17:37:24 <larsbot> you just told me they were

17:37:30 <bijan> But no, they aren't equally expressive.

17:37:34 <bijan> no I didn't.

17:37:43 <larsbot> why are they not? what is RDF missing?

17:38:00 * tim had like danbri realised he had stuff to do

17:38:06 <bijan> Bye tim.

17:38:14 <bijan> It's missing what datalog is missing.

17:38:20 <bijan> Which you didn't understand :)

17:38:24 <larsbot> what is RDF missing?

17:40:19 <bijan> Hmm. Let me try by an analogy.

17:40:40 <bijan> Do you agree that the propositional calculus is strictly less expressive than the predicate calculus?

17:41:02 <larsbot> that's a sidetrack, I think

17:41:11 <larsbot> it's much better if you explain it in terms of RDF

17:41:34 <bijan> How do you know?

17:42:08 <larsbot> because we're talking about data models, of which RDF is an instance, while the calculi are not

17:43:07 <bijan> Again, you're making distinctions that don't exist.

17:43:25 <bijan> I'm not talking about the inference rulse of the propositional calculus.

17:43:35 <bijan> I'm talking about the syntax and semantics.

17:43:49 <larsbot> yeah, but semantics are usually not part of a data model

17:44:05 <larsbot> not beyond "A represents B" and "A contains C"

17:44:06 <bijan> Ok, we're stopping now.

17:44:10 <dmiles> ok you could just say syntax bijan :)

17:44:11 <larsbot> I think that's just as well

17:44:27 <bijan> But, who said I wanted anything *more* from the semantics?

17:44:44 <larsbot> let's drop it; we're filling a channel to no purpose

17:44:51 <larsbot> and I really must go

17:45:06 <dmiles> bijan, could you go on for me?.. I understand you mean what can or cant be expressed syntacticly in predicate/propositional calc

17:45:45 <bijan> Well, Datalog to Prolog is analogous.

17:46:13 <bijan> datalog can express the relational calculus directly

17:46:29 <bijan> I *think* that raw, basic RDF is no stronger than datalog

17:47:43 <dmiles> possibly because datalog/RDF/prolog are all constrained to only a representation

17:48:09 <dmiles> the more meta predicates you define the more expressive you could be

17:48:33 <bijan> Well, look at N3.

17:48:47 <bijan> It adds some stuff.

17:48:54 <bijan> It's more than plain rdf.

17:49:34 <dmiles> because it had to add things to make the represantations usefull

17:51:06 <dmiles> i worry in my boat that SUO-KIF (anohter semantic web effort) will lack the nessisary meta language to be used in the real world

17:51:30 <bijan> KIF is plenty.

17:52:01 <bijan> I'll bet.

17:52:09 <dmiles> well SUO is an effort to ISOize KIF into a processor langauge

17:52:42 <bijan> As long as you allow a full KIF system conformance profile, I have trouble imagining that you'll have much difficulty :)

17:53:00 <bijan> damn. I should have pointed to the kif spec.

17:53:17 <bijan> It specifies stuff equivalent to pure datalog and to a relational system.

17:53:21 <bijan> And explains *all* the terms. :)

17:53:42 <bijan> A Horn system (e.g. pure Datalog) is one in which (1) all assertions are rules that are Horn, unquantified, and baselevel and (2) all queries are positive, non-recursive, unquantified, and baselevel.

17:53:42 <bijan> A relational system is one in which (1) all assertions are rules that are simple, unquantified (but may be non-Horn and non-recursive), and baselevel and (2) all queries are logical, non-recursive, unquantified, and baselevel.

17:54:24 <dmiles> yes

17:55:00 <dmiles> i need to see more of datalog

17:55:16 <bijan> Warren and Maier's

17:55:20 <bijan> Computing with Logic.

17:55:21 <bijan> Get it.

17:55:39 <dmiles> well it just sounds like it has been a standardized and has an inference engine that can run it

17:55:46 <bijan> Worth having anyway.

17:55:51 <dmiles> it might be possible to translate SUO into datalog

17:56:07 <bijan> But they develop "proplog", "datalog", and "prolog"

17:56:27 <bijan> In each case giving an intuitive description, a formal description, an interpreter, etc. etc.

17:56:30 <bijan> Tour de force.

17:56:49 <bijan> And a wonderful book.

17:57:38 <dmiles> ah proplog is fro perl

17:57:44 <dmiles> err for perl

17:59:42 <bijan> Hurm?

18:00:03 <bijan> proplog (in the sense I meant) is "propositional prolog".

18:00:36 <dmiles> yes.. i was just looking to glean code :)

18:00:53 <bijan> it's all pascalish in Computing with Logic.

18:01:01 <bijan> I was thinking of coverting it to python or some such.

18:01:28 <dmiles> i have an inference engine i have been weritting the last year in prolog for work.. and the more sound and complete it gets the slower it goes..

18:01:59 <dmiles> it is a n^3 problem

18:02:03 <bijan> Heh.

18:02:38 <dmiles> so now we have 3 engineers that are ussing it full time.. and reporting bugs.. every bug fix just slows it down more

18:03:03 <dmiles> becaseu it requires a removal of some premature optimization

18:03:18 <bijan> I'll bet.

18:03:21 <bijan> What prolog are you using?

18:03:30 <dmiles> like peoplke want to say (not (isa ?X Formula))

18:03:32 <dmiles> SWI

18:03:38 <bijan> Interesting!

18:03:43 <bijan> Ouch.

18:03:58 <bijan> is the knowledge base just in the prolog one?

18:04:01 <dmiles> and get all bindingings? like millions of answers

18:04:10 <dmiles> no its KIF

18:04:28 <dmiles> i wroite a KIF to prolog compiler

18:04:37 <bijan> No, I mean you compile stuff into the internal db.

18:04:43 <bijan> You aren't using anything external.

18:05:02 <bijan> I.e., assert, retract, etc.

18:05:07 <dmiles> well 3 KBs are generated..

18:05:31 <dmiles> 1st is the surface form.. that javqa makes.. that is sent over a socket to SWI

18:05:51 <dmiles> then SWI canonicalizes and deduces what it can from the KIF forms

18:06:02 <dmiles> and pputs it into two differnt DBs

18:06:21 <dmiles> one DB is for negatin by explicion and the other by failure

18:06:37 <bijan> What I mean is, have you tried tuning the Prolog builtin db.

18:06:47 <dmiles> then it runs a meta interpretor over both DBs for a users query

18:07:00 <bijan> Ouch.

18:07:17 <bijan> And you hook the rules in the meta interpretator? That could get slow.

18:07:20 <dmiles> i was going to start switching th erecorda/ over asserta

18:07:50 <bijan> i mean, have you tried things like index/1, etc.?

18:07:54 <dmiles> start ussing recorded database rather then assert the statemenmtns

18:08:04 <dmiles> yes.. and definately that has helped

18:08:23 <dmiles> i even have been indexing parts of the meta interpretors code

18:08:26 <bijan> Is it all in the queries? How much overhead does the meta-interpreter add?

18:08:57 <dmiles> well i can give you a peek at the meta interpretors code

18:09:34 * bijan in meeting. Email is bparsia@email.unc.edu

18:11:15 <dmiles> sent

18:11:37 <dmiles> that is just one of 40 files

18:11:54 <dmiles> but that one reprents the toploop of interpretation

18:15:41 <dmiles> if i actually wrapped the compiled KIF into prolog asserions it would speed it up 20 times

18:16:07 <dmiles> but its hard to tear down and put back up each time i need to fix a bug

18:17:28 <dmiles> also makes loading a large KB much much slower.. and certain KIF assertions would require a near recompile if someone retracted something that changed something that effected _how_ i compiled in the first place

18:18:15 <dmiles> for example.. if some declares that a certain predicvate is now reflexive.. meanin ghte two arguments could be exchanged..

18:19:04 <dmiles> the compiled version would to be regenerated to allow for that..

18:19:40 <dmiles> so that is why it is never really compiled, it must be interpreted

18:38:38 <SethR> dmiles, what does (not (isa ?X Formula)) mean ?

18:40:30 <dmiles> to find all instances 'disjoint' with 'Formula' class and that maybe should not look like formula's syntactically as well

18:41:34 <dmiles> so the method of finding that would be to first collect disjoint classes to 'Formula'

18:42:00 <dmiles> which would be everuything under 'Oject'

18:42:45 <SethR> so it picks out the nodes as oppesed to the arrows?

18:43:00 <SethR> oppesed= vs

18:43:25 <dmiles> yes,, the instances of the disjoint classes instead of the clases themselves

18:44:08 <dmiles> which this could mean everything known that is not a formula in the entire world

18:44:59 <dmiles> i used 'isa' instead of 'instance' since there are more cyc people ouyt there

18:45:03 <SethR> so objects in the world *as well as* the nodes that represent them

18:46:03 <dmiles> yes.. in that case everything

18:46:32 <SethR> everything except the arrows themselves which are the Formula

18:46:37 <dmiles> but only things that we could 'prove' to not be instances of Formula

18:47:20 <dmiles> not things we have 'failed to prove' as Formula

18:48:11 <dmiles> so could we prove the arrow to not be part of Formula?

18:48:28 <dmiles> err not Formula

18:49:21 <dmiles> the arrows represnt a relationship of classes right?

18:49:33 <dmiles> or do you mean entailment?

18:50:13 <SethR> arrows are relationships between nodes representing relationships between objects in other contexts

18:52:37 <SethR> im just starting to grok entailment

18:53:05 <SethR> would you say:

18:55:13 <SethR> {?p => ?q} |= [or ?~p ?q] ?

18:56:19 <SethR> {?p => ?q} |= [or ?~p ?q]. ?~p dmal:complementOf ?p.

18:57:59 <SethR> in other words change dmal:equivalentTo to |= entails in http://robustai.net/mentography/implies.gif

19:00:46 <SethR> or should it be |- meaning just that it is provable?

19:02:13 <SethR> should it read {?p => ?q} |- {[or ?~p ?q]. ?~p dmal:complementOf ?p.}

19:02:35 <AaronSw> Hmm, someone called me while I was out but I can't figure out who it was...

19:02:38 <AaronSw> hmmpg

19:02:56 <SethR> hi aaron

19:03:04 <AaronSw> hi seth

19:03:31 <SethR> does n3 have the entailment operator ?

19:03:48 <AaronSw> not to my knowledge

19:31:23 * tim tunes in

19:33:13 <tim> If n3 were to have an entailment operator, then would it be used to assert that one formula entails another - introducing a rule - or would it be used as a binary operator to query whether one formula, under deduction, leads to another?

19:33:22 <tim> Or could the language have th esame symbol for both?

19:33:36 <tim> Theer is log:implies for teh first one.

19:34:20 <tim> There is log:includes as a simplistic version of the seodn one, where log:includes does not use any axioms and only does trivial inference.

19:49:17 * DanC tunes in...

19:50:05 <bijan> hey DanC.

19:53:12 * DanC has a telcon in 7 minutes

19:53:47 <DanC> re implies vs. entails and such: I think perhaps I have grokked that, for very short periods of time. But I can't seem to keep it in my head.

19:53:57 <bijan> oh no! :)

20:11:02 <tim> (Basically, if you can define what it does then I can code it up, if you can't it doesn't worry me!)

20:15:24 <tim> DanC, if over the weekend a simple statement of the differenece between implies and/or includes and entails surfaces by osmosis, then I would be intereted in a tutorial.

20:16:10 <bijan> Tim: the only diff was that (often) entails is use for semantic inference whereas implies is for syntactic.

20:16:17 <bijan> I.e., |= vs. |-

20:16:55 <bijan> Which you use in a sound and complete system is, well, largely moot :)

20:18:32 <bijan> If I understood your proposal, the first one would allow the introduction of unsound rules?

20:22:59 <tim> Ah yes, I vaguely remember that basically we are talking about two systems, |- in one and |= in the other, and soundness and completeness being the subset/equality/superset status of the comparison of the |* in each system.

20:24:13 <tim> one of the systems being something one refes to as realty or the underlying semantics or interpretation or the model theory -- annother abstract system but one which has some faith in.

20:25:53 <tim> Sigh ... every time I read this stuff I sem to be waterproof to absorbing the conventional use of terms.

20:40:09 <SethR> i was getting the drift that |* operators were about logic, and the => operators are about things .. but i might have been getting that wrong

20:42:09 <SethR> being about logic = being about rules for substituting statements for statements

20:43:08 <SethR> being about things = being rules that restrain the world of the domain of discourse

20:44:10 <SethR> does that make sense to anybody, or am i just confused?

20:50:31 <bijan> Tim, it really helps to work through an intro logic book now and again. Slam the terms home :)

20:51:30 <bijan> As for the "referring to reality" vs. "abstract system"...er...I don't know :) |= is usually specified by a model theory (though there are inferential semantics)

20:52:24 <bijan> Well, actually, I guess when someone says "model theory" they usually mean explicating |=

20:52:33 <bijan> Proof theory for |-

20:52:47 <bijan> The two can come apart in interesting ways.

21:02:56 <SethR> hey, back to John and Mary ... why can't we use dmal:complementOf between properties

21:03:13 <SethR> {loves dmal:complementOf ~loves}

21:03:43 <SethR> therefore it would be legit to write: {John ~loves Mary}

21:03:49 <SethR> huh?

21:05:00 <SethR> dmal = daml

21:06:53 <SethR> woops! scratch that .. daml:complementOf is everything but

21:14:15 <tim> ComplementOf is sets, not properties.

21:15:01 <tim> But it is true that while p|~p is not in general true of a formula, it is true of some properties (or all formulae which are of the form { s p o } and p is one of a certain set).

21:15:41 <tim> So for example the builtins in cwm string:greaterThan and so on all have ngeated forms, as does log:includes.

21:16:32 <tim> it might be worth noting properties for which this is the case, so that FOL can be used on them?

21:17:37 <SethR> so why can't we have a general ~ property to apply between properties ?

21:18:06 <sandro> Tim, when you asked about blindfold parsing makefiles, were you thinking in terms of having the Make variables expanded before conversion, or remaining as distinct objects?

21:19:03 <tim> General "~" property: Yes, although I wouldn't call it "not". A forumla can be negated -- need term for a property.

21:20:10 <SethR> {~loves ~ loves} so ~ lets me make any Property that denies some other property ... or maybe {~loves denies loves}

21:20:25 <tim> Sandro, I didn't think much. As they can be expanded right in the middle of a filename, I would suggest asimple version expands them and then just talks about filenames whole.

21:20:43 <tim> for make, wed need basename and suffix properties to make the rules.

21:21:45 <SethR> so we could have complementOf for classes and denies for properties

21:22:11 <sandro> So a makefile would parse to several of { fileName isBuiltFrom [ list of file ]; isBuiltBy [ list of shell commands ] }

21:22:35 <tim> [ make:insuffix ".dot"; make:outsuffix ".svg" ; make:recipe ( "cwm foo.doraskasdajsh" "xslt sadhjsdfuhkj > $@" ) ]

21:23:13 * DanC is interested in SemanticWebMake; but every time I start to work on it, I feel obliged to go study Ant (the java/XML build thingy from Apache world) first

21:23:23 <tim> Yes, makefile would parse to that

21:24:29 <tim> What would one need to actually build make? The ability to call a system command -- disable in agents! -- and do subsitutions

21:25:47 <SethR> tim, do we have a 'denies' or some such thing in N3 ?

21:26:54 <tim> log:notIncludes ?

21:27:12 <tim> realtionship between two formula that one doesn't include the information in the other.

21:27:28 <tim> what do you mean by 'denies"?

21:27:40 <tim> "says not"?

21:28:14 <SethR> yes denies would negate a property .... {John (not loves) Mary}

21:29:41 <DanC> tim, what do you use to represent inconsistencies? or rather: is log:Falsehood something you like these days?

21:29:54 <tim> Seth's meaning of denies I think doesn't match my feeling about what the english menas ie says not.

21:29:55 <DanC> I think the traditional term for "~" is converse.

21:30:34 <tim> oh, so log:includes danc:converse log:notInvludes ?

21:30:58 <tim> What namespace should it gointo? WebOnt?

21:31:20 <DanC> so... { this log:forAll :s, :p, :o, :F. { :F log:includes { :s :p :o. :s [is :converse of :p] :o} } log:implies { :F a log:Falsehood } }.

21:31:55 <DanC> ^ that's a question/proposal/suggestion. Does it look right?

21:32:02 <SethR> do we already have :converse ??

21:32:09 <tim> yes. Looks good. In that I know what it means.

21:32:10 <DanC> I'm making up :converse right now.

21:32:30 <tim> At the moment log:Falsehood isn't used AFAIK.

21:32:35 <DanC> do you endorse/like/use/recommend log:Falsehood, TimBL? are there any test cases involving it?

21:33:07 <tim> No, I don't have any test cases. I think you were asking for some HCF instrcution fo cwm

21:33:11 <tim> (Halt and Catch Fire)

21:33:25 <tim> Like assert.

21:33:26 <DanC> was I? I hereby retract any such request.

21:33:54 <tim> You asked if there was anything "intersting" which you could make cwm do it it found something was a member of 2 disjoint classes.

21:34:14 <tim> i assumed you mean thouw an exception.

21:34:15 <DanC> yes, I'm still interested in that.

21:34:27 <DanC> hmm...

21:34:37 * DanC takes a call

21:35:48 <SethR> isnt it ingteresting to use disjoint classess to state rules ?

21:36:24 <tim> {:F log:includes { :s [is :converse of :p] :o }} log:implies { :F log:notIncludes { :s :p :o}}. <- an axiom for inference from :converse

21:36:52 <tim> {:F a log:Truth; log:includes { :s [is :converse of :p] :o }} log:implies { :F log:notIncludes { :s :p :o}}. <- an axiom for inference from :converse

21:37:19 * sbp wonders if you need the "is x of" for a reversible property such as converse

21:39:34 <tim> Oh that's true. :converse dpo:inverse :converse.

21:40:10 <DanC> inverseOf

21:40:11 <SethR> ie its symmetric

21:40:17 <sbp> { ?y a :ReversibleProperty . } log:implies { ?y daml:inverseOf ?y } .

21:40:31 <DanC> s/ReversibleProperty/SymmetricProperty/.

21:40:32 <tim> { this log:forAll :s, :p, :o, :F. {:s :p :o. :F log:includes {:s [:converse :p] :o} } log:implies { :F a log:Falsehood } }.

21:40:49 * sbp notes Seth/DanC's correction

21:41:21 <sbp> use cases for log:Falsehood: I picked up "OR" from the Toolbox

21:42:20 <sbp> e.g. to say "Bob is a Fish, or Bob is a Dog": { { :Bob a :Fish } a log:Falsehood . { :Bob a :Dog } a log:Falsehood } a log:Falsehood .

21:42:23 <tim> { this log:forAll :s, :p, :o, :F. {:s :p :o. :s [:converse :p] :o} log:implies { :p error:Inconsistency "Converse and proerty both apply" }.

21:42:41 <tim> OR? Operational Research?

21:42:52 <DanC> OR as in "this or that".

21:42:57 <sbp> yeah

21:43:08 <sbp> logical OR

21:43:31 <tim> ooops

21:43:49 <tim> Toolbox a :rusty;

21:44:13 <tim> doc:supercededBy :n3.

21:44:44 <jah-home> logical OR = NLP:and/or

21:44:52 <SethR> tim, why the "Inconsistency ... ?

21:45:47 <sbp> how would you express OR now? I thought of using an Alt bag (I thought was was what it was meant to do), e.g. [ a rdf:Alt; rdf:_1 { :Bob a :Fish }; rdf:_2 { :Bob a :Dog } ] .

21:46:59 <tim> I would say something like ( :F :G :H ) a :TrueDisjunction.

21:47:08 <tim> I would say something like ( :F :G :H ) a :OR.

21:47:52 * sbp makes a note

21:47:57 <SethR> where :F :G and :H are statements ?

21:48:09 * sbp interpreted those as statement sets

21:48:13 <tim> {[ a :OR; daml:first [a log:Falsehood]; daml:rest:x]} log:implies {:x a log:Truth} maybe

21:48:47 <tim> statement sets - yes i call them formulae at the moment - get sme into least trouble ;-)

21:49:10 <sbp> heh!

21:49:24 <SethR> i call them arrows :)

21:50:33 <SethR> to me you got arrows or nodes ... that's all ... nothing else

21:50:43 <sbp> ABC - Anything But Contexts

21:51:20 <SethR> ok ok ... arrows ... nodes ... and contexts

21:53:09 <SethR> but really all you got is arrows .. cause the arrows creats the nodes and the contexts

21:53:51 <sbp> Gotta run

21:53:57 <SethR> resources being things outside the system that the nodes or the contexts represent

21:54:03 <SethR> by sean

21:54:40 <jah-home> formulae, interesting name :->

21:55:09 <SethR> that's what SUO calls them

21:56:51 <jah-home> I wonder if it wouldn't be worth writing a document somewhere with some of these logical "idioms" defined in a way we could use FOL:OR, FOL:AND, etc - would allow someone to build a parser to take things in more traditional form (but still URIed) and turn into RDF easily. Can't decide if I think this would be a step forward or backward now that I'm getting to like n3/cwm syntax

21:59:08 <SethR> you mean like kif

22:02:20 <tim> A formula is not an arrow. A statement is an arrow. A formula is equivalent to a relational database.

22:04:10 <jah-home> SethR - like KIF if it was made web friendly (and fixed fr a lot of ugly math that isn't needed)

22:06:29 <tim> Just addingnamesapces would be a good start.

22:06:42 <tim> s// / as appropriate

22:08:04 * DanC updates webont WG membership records, frustrated that it isn't very automated yet

22:11:58 * tim tries to find bug in new supposedly faster hashing in llyn.py query

22:18:44 <DanC> tim, I use 2001/05ve/rdfs.n3 for RDFS rules (domain/range, subclass, etc.); do you have those rules written somewhere else?

22:18:58 <DanC> it's not the most natural home for them

22:27:36 <DanC> hm... does cwm's string-handling have enough oomph to say "email addresses are case insensitive"?

22:27:39 <DanC> e.g. [[

22:27:40 <DanC> <mailbox rdf:resource="mailto:Lynne.Thompson@unisys.com"/>

22:27:40 <DanC> <mailbox rdf:resource="mailto:lynne.thompson@unisys.com"/>

22:27:41 <DanC> ]]

22:28:52 <DanC> base on { "mailto:Lynne.Thompson@unisys.com" string:equalIgnoreCase "mailto:lynne.thompson@unisys.com" } and so on, I'd like it to conclude those mailboxes are daml:equivalent, and by way of forgetDups.n3, that one of them is Chaff

22:30:08 <DanC> { :m1 a contact:Mailbox. :m2 a contact:Mailbox. :m1 log:uri [ string:equalIgnoreCase [ is log:uri of :m2]]. } log:implies { :m1 = :m2 }.

22:30:09 <tim> Aaron gave me a list of pointers one to all the places those rules were written up.

22:30:38 * DanC infers that "those rules" refers to RDFS rules, and thinks the pointer to was to RDF tidy...

22:31:02 <DanC> rather, RDF lint...

22:31:10 <tim> I don't have a definitive set. I use only something in a primitive schema validation example.

22:31:13 <DanC> ... http://infomesh.net/2001/05/sw/#rdflint

22:31:21 <tim> Jos has a bunch.

22:31:47 <DanC> ... http://www.w3.org/2001/03/earl/srules-lite.n3

22:32:06 <DanC> string:equalIgnoreCase isn't there (yet) is it?

22:32:09 <tim> I would suggest string:upper and string:lower as new builtins.

22:32:53 <DanC> my experience says case-insensitive-comparison is the right idiom, not (information-destroying) string:upper/lower

22:33:54 <tim> Not [ string:upper :x ] log:equalTo [ string:upper :y ] ??

22:34:00 <DanC> nope.

22:34:22 <DanC> it necessarily involves going over the length of each string

22:34:30 <tim> Well, then you need isuppose > and >= and = != and < and <= ?

22:34:39 <DanC> where as string:equalIgnoreCase might win by just looking at one character of each string.

22:34:58 <tim> From the performance point of view, certainly.

22:35:34 <DanC> and from experience with filesystems: case-preserving-but-case-insensitive is the right answer.

22:35:57 <DanC> you almost never really want stomp on the original string.

22:37:10 <tim> You mean, give someone a rope to convert stuff to upper case and they'll hang us all with it.

22:37:10 <tim> We will ose the infromation. yes.

22:38:11 <DanC> I'm actually uneasy about the "email addresses are case-insensitive" rule. Strictly speaking, they're not. domain-names are, but the email specs don't license the inference that <mailto:DanC@w3.org> = <mailto:danc@w3.org>.

22:38:34 <DanC> the RFCs do say <mailto:danc@W3.ORG> = <mailto:danc@w3.org>

22:39:28 <DanC> Jim Whitehead just claimed, in a mail message to uri@w3.org, that URI scheme names are case-insensitive. That gave me the creeps.

22:41:30 <AaronSw> they're not?

22:42:07 <DanC> The RFC says "for resiliency, they should be compared without regard to case" or words to that effect. i.e. "to account for bugs, ..."

22:42:43 <AaronSw> Hmm.

22:42:43 <DanC> the reason it gives me the creeps is that I don't want folks to expect URI-consuming-gizmos to do anything other than strcmp()

22:43:35 <^lilo> [GlobalNotice] Hi all. We are presently implementing 10-user-per-IP limits, with a few exceptions. If you need an exception, let us know. If you run into problems connecting to the network, let us know that too. Email can be sent to support@openprojects.net. Thanks!

22:43:53 <DanC> e.g. cwm isn't going to notice that <http://example> = <HTTP://example/>

22:44:01 <DanC> unless you tell it so.

22:44:02 <AaronSw> hmm

22:44:05 <AaronSw> makes sense

22:44:16 <tim> Absolutely.

22:44:20 * jah-home catches up on the log - now understands why his students are having so much trouble finding some of the cwm and rule stuff (raiding cvs repositories is current method) - perhaps W3 sw, or else AaronSW could create a "definitive" page in non-member space with appropriate pointers? Perhaps such exists and I've just missed it?

22:44:30 <tim> Hwoever, if you want to feel less creepy ...

22:44:45 <AaronSw> raiding CVS repositories is the Right Thing

22:45:01 <tim> Consoder that with any two URIs the may be some great similary or form of identity, which you can find if you use more or less knowledge.

22:45:06 * DanC hopes the DRUMS WG decides that mailbox names are case-insensitive anyway, since so many implementations assume that.

22:45:15 <AaronSw> DRUMS?

22:45:36 <DanC> the IETF WG named "DRUMS". I trust you can find it.

22:46:08 <DanC> Detailed Revision/Update of Message Standards (drums) Charter http://www.ietf.org/html.charters/drums-charter.html

22:46:43 <DanC> the WG that obsoleted RFC822

22:47:01 <tim> JimH, which cvs repository were your folks raiding? cvs.w3.org?

22:47:03 <DanC> phpht. 404 in the path from their home page to RFC2822

22:47:16 <DanC> tim, let's clean up this "how do I get the cwm" code right now...

22:47:24 <AaronSw> There are going to be more 822 RFCS? eeek

22:47:38 <mnot> already been ;)

22:47:50 <tim> Can we capture the Error 404 there as an example of why I wan the IETF to make a persistent store?

22:48:07 <DanC> Tim, how do you expect folks to find the swap stuff from the W3C home page? via the Semantic Web page, then via SemWeb AD?

22:48:41 <tim> I suppsoe so. Also from toation3 and DesignIssues.

22:48:42 <DanC> oops... "swap" doesn't occur in http://www.w3.org/2000/01/sw/

22:49:06 * DanC wonders how folks find DesignIssues from W3C home

22:49:11 <tim> What *is* the public access recipe?

22:49:23 <jah-home> tim - yes cvs.w3.org -- DanC clean up now - YAY!

22:50:12 <DanC> public read access to SWAP code is via dev.w3.org ; exported into HTTP space at http://dev.w3.org/cvsweb/2000/10/swap/

22:50:14 <jah-home> re - designissues, I find via google when I don't have a bookmark handy - there is a path through guide, but need a member password to go that way, i did find a handy path once, but lost it ...

22:50:33 <DanC> command-line CVS instructions at http://dev.w3.org/cvsweb/

22:50:46 <AaronSw> mnot, more than 2?

22:50:59 <mnot> my biggest problem these days is figuring out what cwm is capable of re: n3, rules, etc.; there's a lot of dissonance between some of the examples and the code

22:51:01 <DanC> tim, I'd do the update of the SWAP page for you, but that won't make sure it's in your head.

22:51:02 <mnot> er...

22:51:18 <lilo> [GlobalNotice] Hi all. Just a further note. The connections-per-IP limits are not meant to be onerous, but we've had problems with clone attacks. Thanks.

22:52:37 <mnot> Well, there are the ever-cute rfc2821 and rfc2822 (http://rfc2821.x42.com/ http://rfc2822.x42.com/)

22:53:05 * DanC requests a pause in the channel while we update the SWAP page to make clear how to get the code from CVS

22:53:26 <AaronSw> yah, thanks

22:53:35 <DanC> tim?

22:53:51 * DanC does it for him...

22:53:58 <tim> o-oh

22:54:08 <tim> Document saved

22:54:11 <DanC> you wanna do it, tim?

22:54:17 <mnot> looks like they haven't updated their milestones in a while,

22:54:32 * DanC finds 1.18

22:54:53 <DanC> pls point to http://dev.w3.org/cvsweb/2000/10/swap/ as well as http://dev.w3.org/cvsweb/ , tim

22:54:59 <tim> ^R

22:55:12 <tim> I had done it ... it just takes a while to save.

22:55:26 <tim> 1.189 does

22:55:30 <tim> 1.19 does

22:55:36 <DanC> "the w3c CVS

22:55:36 <DanC> repository.

22:55:36 <DanC> " is an ill-formed definite description; there are two.

22:56:03 <DanC> it calls itself "the W3C public CVS tre

22:56:04 <DanC> "

22:56:05 <tim> public W3C CVS repository

22:56:45 <DanC> JimH, pls review http://www.w3.org/2000/10/swap/ v 1.19 . AaronSw too.

22:57:40 <tim> That and a circles-and-arrows diagram derived automatically and folks should understand everything.

22:58:00 <AaronSw> it'd be cool if you note by the direct links to stuff like cwm.py that they need to grab the source, but this is good

22:58:00 <tim> (no makefile, though. Dependencies are rather hiffen.)

22:58:04 <DanC> tim, pls s/-/--/g in the command-line doc section, while you're there?

22:58:22 <lilo> [GlobalNotice] Hehe, this is, shall we say, a whole can of worms. 8) Perhaps we should try to find another answer to the clone problem. ;)

22:58:42 <DanC> also, make note of the tests-are-the-truth policy, and point to retest.sh?

22:58:47 <AaronSw> Also cool would be to put something like sf has...

22:58:47 <AaronSw> cvs login -d ":pserver:anonymous@dev.w3.org:/sources/public

22:58:47 <AaronSw> "

22:58:47 <AaronSw> etc.

22:59:05 * AaronSw gets the details

22:59:10 <DanC> AaronSw, the cvs login -d prayer is linked.

22:59:22 <AaronSw> oh?

22:59:35 <AaronSw> i don't see it? hint?

22:59:37 <DanC> prayer is atop http://dev.w3.org/cvsweb/

22:59:40 <AaronSw> nope

22:59:43 <AaronSw> just the id

22:59:45 <deltab> I've found that I can do checkouts without login, at least with some repositories

22:59:53 <AaronSw> Hmm.

23:00:03 <AaronSw> deltab, perhaps you were already logged in?

23:00:04 <DanC> ah... prayer is insufficiently clear. I get it.

23:00:11 <AaronSw> yeah.

23:00:21 <gerald> http://validator.w3.org/source/ may be clearer

23:00:21 <dc_rdfig> F: http://validator.w3.org/source/ from gerald

23:00:26 <gerald> doh

23:00:33 <deltab> no, it happened with a project I'd not heard of before

23:00:35 <AaronSw> hey gerald

23:01:11 <AaronSw> yeah, the validator version is clearer

23:01:37 <DanC> gerald, do you know how to edit http://dev.w3.org/cvsweb/ ?

23:01:51 <gerald> not offhand... maybe a README.cvs somewhere?

23:02:01 <tim> We have guests ... better go!!!

23:02:14 <DanC> ok, tim, ciao. v1.19 is latest?

23:02:17 * tim realeases write lock

23:02:24 <tim> tim is now known as tim-lurk

23:02:27 <DanC> release lock at v1.19?

23:03:06 * DanC finds v 1.21

23:03:21 <DanC> v 1.21 includes testing policy and s/-/--/.

23:03:22 * DanC is happy

23:03:28 <mnot> I don't suppose there's anything documenting the overall architecture of cwm / llyn / etc.? *crosses fingers*

23:03:38 <DanC> ok, you can have the channel back.

23:04:05 <DanC> err... in a word, mnot, no.

23:04:35 <mnot> ah, well, I guess I know what I'm doing on the weekend ...

23:04:44 <deltab> cvs checkout: used empty password; try "cvs login" with a real password

23:05:01 <DanC> the architecture is documented, but understanding it is a gestalt experience one gets after swimming around the tests and the /DesignIssues/ pages for... oh... say... 4 years.

23:05:09 <AaronSw> lol

23:05:50 <mnot> weekend may be ambitious, then.

23:05:54 <DanC> ;-)

23:06:54 <mnot> AaronSw: uri stuff going well; this works:

23:06:57 <mnot> this log:forAll :x, :y, :z, :a, :b, :c .

23:06:57 <mnot> { :x :y :z .

23:06:57 <mnot> :x log:uri [ uri:scheme :a ] }

23:06:57 <mnot> log:implies

23:06:57 <mnot> { :x x:theScheme :a } .

23:07:06 <AaronSw> cool

23:07:09 <mnot> :)

23:07:31 <AaronSw> oh, heheh

23:07:57 <AaronSw> hi edd

23:07:58 <DanC> you guys are formalizing URI syntax in n3 rules?

23:08:04 <mnot> bit by bit

23:08:06 <AaronSw> edd, missed you at p2pcon

23:08:13 <AaronSw> :-(

23:08:19 <mnot> I'm reapproaching URISpace from the other direction

23:08:20 <edd> hey AaronSw. p2pcon -- too overhyped for my tastes

23:08:21 <DanC> I started doing that (formalizing URI syntax) in larch a while back; perhaps we should compare notes.

23:08:36 <AaronSw> overhyped? everyone else complained it was underhyped...

23:08:46 <DanC> hi edd. been a while. happy hacking?

23:08:56 <mnot> Yeah, Tim pointed me at that. I was afraid I'd blow my stack if I tried to learn rdfs, daml *and* larch at the same time...

23:09:22 <edd> DanC: if only... Real Life been intervening too much.

23:09:30 <DanC> I see, edd.

23:09:43 <DanC> well, mnot, I can compare notes then.

23:10:00 <mnot> heh. OK, I'll send along a reference when I've got a bit more together.

23:10:36 <mnot> hints about hacking cwm to support it would be, of course, *very* appreciated.

23:10:37 <DanC> so... I gather uri:scheme works ala { "http://example/" uri:scheme "http" }?

23:10:45 <mnot> yes

23:10:51 <mnot> uri:path is a list

23:11:09 <DanC> hmm.. you might consider { "http://example/" uri:scheme [ uri:schemeNamed "http"] }.

23:12:02 <DanC> or rather urilist:schemeName , where @prefix urilist: <http://www.w3.org/Addressing/schemes#>.

23:12:06 <mnot> yes. Taking baby steps first.

23:12:19 <mnot> ah, interesting

23:13:55 * mnot sees http://www.w3.org/Addressing/schemes.n3 - perfect

23:14:51 <DanC> yeah; had fun with that; I gather you went via http://www.w3.org/Addressing/schemes.html#ctech ?

23:15:21 <mnot> yep

23:15:34 * AaronSw had fun with that too

23:15:44 <AaronSw> altho i remain afraid of the xslt

23:16:10 <DanC> TimBL has been working on making cwm powerful enough to cut xslt out of the loop in the case of rdf2dot.xsl

23:16:29 <DanC> his techniques should work for generating XHTML, eventually, too.

23:16:29 <AaronSw> cool

23:18:06 * DanC is jazzed at this sort of approach to URISpaces; did we point in this direction in the staff comment on it?

23:18:39 <mnot> not really; just wasn't to familiar with the approach of RDF, etc. at the time. Discussions with danbri in HK helped.

23:18:56 <DanC> HK?

23:19:00 <mnot> Hong Kong

23:19:04 <DanC> ah

23:19:37 * mnot wants to build a Web server which is completely rule-driven

23:19:39 <DanC> actually, the staff comment does (very, very subtly) point in this direction.

23:19:48 * mnot looks back...

23:19:48 <DanC> via aboutEachPrefix

23:19:53 <DanC> cf http://www.w3.org/Submission/2001/02/Comment

23:20:16 <mnot> ah, yes. aboutEachPrefix in all of its glory

23:20:56 <dajobe> oops, did we prematurely delete it?

23:21:18 <AaronSw> no

23:21:19 <DanC> no; mnot is busy implementing it propertly.

23:21:27 <dajobe> phew!

23:21:30 <mnot> heh

23:21:52 <DanC> i.e. layering it on top of rules and the magic log:uri level-breaker.

23:22:03 <dajobe> yeah

23:23:41 <dajobe> which is where aboutEach should live; iff we have sufficient reasons to remove it

23:24:41 <DanC> it not only should live there; it does: running code: http://www.w3.org/2001/04rs22/#ae

23:27:00 <dajobe> yeah I knew of this. Just zapping aboutEach is just rather at the boundary of syntax bugfixes and needs care

23:28:31 <jah-home> JimH leaves to go finish cooking the turkey and eat his early Tgiving meal -- happy holidays everyone!

23:34:22 <DanC> yes, very much at the boundary. Pretty clearly over the boundary, without the aid of reality-distortion goggles.

23:34:55 * DanC is getting resigned to the fact that the RDF Core charter needs changing. We're doing RDF 1.1.

23:35:41 <dajobe> well...

23:35:47 * edd raises an eyebrow

23:36:06 <DanC> or 1.0.1, to follow the HTML 4.0.1 precedent.

23:36:47 <AaronSw> Heh.

23:36:48 <DanC> i.e. these are real changes. Not changes that anyone in their right mind would protest, but technically observable changes nonetheless.

23:36:58 <mnot> from my somewhat-outsider perspective, I'd say so; comments like "RDF containers are broken" in reference to a Recommendation tend to imply more than a cleanup, IMHO

23:39:05 * dajobe picks a number <1 out of the air, adds it to 1.0

23:40:14 <dajobe> I never believed the Errata bits only would work; skeptical about RDF 1.0 2nd Edition idea

23:41:31 * AaronSw really hopes next WG is chartered to do 2.0

23:41:44 <AaronSw> And will be very upset if it is not... *nudge*

23:41:52 <dajobe> yeah, in the next one

23:42:23 <DanC> hmm... RDF 2.0 brainstorm... just a new syntax for the same expressive capability?

23:42:24 <dajobe> plus standardising non-XML syntaxes for RDF

23:42:36 <AaronSw> DanC, no... more expressive

23:42:47 <AaronSw> i'm not interested in a syntax WG

23:42:47 <dajobe> . rdf2:isNot

23:42:48 <DanC> or should RDF 2.0 include stuff like quoting, log:uri, log:implies maybe even?

23:42:58 <AaronSw> yes

23:43:02 <AaronSw> definitely quoting

23:43:05 <AaronSw> not sure about the rest of it

23:43:16 <AaronSw> and new containers

23:43:23 <AaronSw> and contexts

23:43:25 <dajobe> I feel n3 shows strong hints what to add; what is used most

23:43:25 <AaronSw> that sort of thing

23:44:10 <DanC> N3 is darned useful, but I have to say that it's *very* non-traditional, from a logical perspective. Its complexity (read: performance) characteristics are anybody's guess.

23:44:54 <DanC> new containers: there's an interesting tension-point. Does RSS rely on RDF 1.0's container vocabulary?

23:45:35 <DanC> darned useful: as long as your files are fairly small ;-)

23:45:37 <AaronSw> It uses it...

23:45:42 <AaronSw> RSS, that is

23:46:12 * edd admits it was his fault :)

23:46:30 <AaronSw> Heh.

23:46:35 <DanC> well, W3C Recommended the squirrelly things, so it's not all your fault, edd.

23:46:52 <AaronSw> It also confuses properties and classes, tho...

23:47:37 <edd> as well as confusing the Frontier-wielding populace

23:47:43 <edd> oh what havoc we wreaked

23:48:13 <AaronSw> Heh heh heh.

23:48:23 <DanC> I don't suppose many of the RSS APIs have an RDF-ish makeStatement feel to them, do they?

23:48:40 <AaronSw> makeStatement? no

23:48:48 <AaronSw> Well...

23:48:53 <AaronSw> Orchard is close

23:49:00 <DanC> makeStatement as in http://www.w3.org/2000/10/swap/RDFSink.py

23:49:02 <AaronSw> you can say item.title = 'bar'

23:49:26 <AaronSw> generates :item :title "bar" .

23:49:52 <DanC> that's pretty cool.

23:50:40 <edd> i wish more did.

23:50:55 <AaronSw> I'm planning on doing that for my rdfapi

23:51:03 <DanC> any hope of moving the Frontier crowd in that direction?

23:51:04 <edd> saddling RSS 0.91 with a DTD was the worst idea ever

23:51:24 <DanC> DTDs... yes... another fine W3C product. phpht.

23:51:49 <sandro> I clever idea pushed out the door before it was really wellunderstood....?

23:52:15 <sandro> s/I/A/

23:52:37 <DanC> er... no, it escaped out the door under 10 years of momentum. It was widely known as a dying beast, but it lived long enough to get out the door.

23:57:02 <sandro> I'm curious to see how well my mapping BNF & XML-Schema to the same grammar ontology helps me also map to/from DTDs. (if I ever have a good reason to go there)