Semantic Web Interest Group IRC Chat Logs for 2003-03-25

04:27:18 * DanC noodles on where to fit aboutEachPrefix into ESW faq...

04:42:41 <DanC> hmm... it's not an exception to URIOpacity as such...

04:43:26 <DanC> it's more a case of less information hiding than usual...

04:44:04 <DanC> BreakingURIOpacity isn't exactly right, but it seems to be the best name I can come up with

05:13:40 <DanC> bummer... no blockquote support in TWiki

05:21:35 * DanC saves 1st part of http://esw.w3.org/t/view/ESW/BreakingURIOpacity

06:46:26 <taaz> taaz is now known as taazzzz

09:36:34 <golbeck> golbeck is now known as golbeck_zzz

14:02:51 * libby updates calendar testcases to use timezones.

14:03:24 <bitsko> is anybody maintaining an RDF calendar on their webpages yet?

14:04:01 <bitsko> I recall a little while back there were some toe-dippings in that area, doing itineraries and such

14:04:06 <libby> heh, well the RDF interest group calendar taskforce is.....

14:04:42 <danbri> That's a good question. I'm not, personally. Need to set aside the sense that I "should" be doing it and find a reason to make it useful for me to do it...

14:05:25 * libby has an idea

14:05:29 <libby> re good reasons

14:05:36 <danbri> yeah?

14:05:38 <libby> well, a sort of good reason

14:06:02 <bitsko> I run a news site (er, or did until the server crashed) that would make use of an "upcoming events" calendar

14:06:32 <libby> so if it was easy to search rdf calendar files and generate a schedule from existing events and merge them with your own, then maybe that would eb an incentive...

14:07:29 <libby> there are some rss events files around, but the new schema woudl be better. need to chat to those who have feed.

14:07:37 <bitsko> and one side interest I have in a calender application is that it merge timestamped but not-calendared events into the calendar (think publications, webloged items, email threads, bills, ...)

14:08:37 <danbri> yup, a *lot* of things are timestamped...

14:09:24 <dajobe> logger_1, where am I?

14:09:24 <dajobe> See http://ilrt.org/discovery/chatlogs/rdfig/2003-03-25#T14-09-24

14:09:52 <danbri> logger_1, how did I get here?

14:10:22 <logger_1> I'm logging. I found 1 answer for 'how did I get here'

14:10:23 <logger_1> 0) 2003-03-25 14:09:52 <danbri> logger_1, how did I get here?

14:10:58 <dajobe> probably a bad idea

14:10:58 <dajobe> as it goes to search years of chat logs

14:10:58 <danbri> oh, i thought the keyword 'grep' was needed

14:10:58 <dajobe> nope

14:18:57 <bitsko> DJ Adams (qmacro) has blosxom set up to deliver RDF from either the index (rss:channel) or an item (rss:item). I was thinking on building sideways from that reference other items (foaf, software, calendar)

14:21:22 <danbri> i'm hearing more and more interesting things re blosxom lately, i sohuld revisit it

14:21:26 <danbri> that sounds like a good project

14:21:30 <danbri> its v hackable, i hear...

14:21:37 <bitsko> one question that immediately surfaced has to do with contexts. I know a lot of work is going on in contexts, but specifically, how does one prevent creating additional triples for resources that have been effectively read already?

14:21:56 <bitsko> Blosxom 2.0 is very hackable, with lots of good community plugins already

14:23:04 <bitsko> and dang quick to set up. "under 15 minutes or your money back" ;)

14:23:44 <libby> bitsko, that sounds great

14:23:51 <bitsko> - http://www.raelity.org/apps/blosxom/downloads/blosxom_2_0_rc2.zip

14:24:36 <bitsko> that contains one file, the blosxom cgi. put it where a .cgi can run, edit the top of the file to point to a directory you can write in, go edit your '<dir>/first-post.txt'. done.

14:25:01 <libby> if you do get the urge to do calendar stuff, there's lots of info and test data here: http://www.w3.org/2002/12/cal/ (esp in /test), and I'm happy to help if you run into problems

14:25:38 <bitsko> all files '*.txt' in <dir> are weblog entries, other files (that can come later) are usually templates

14:50:53 <bitsko>http://stx.sourceforge.net/

14:50:54 <dc_rdfig> A: http://stx.sourceforge.net/ from bitsko

14:51:08 <bitsko> A:|Streaming Transformations for XML (STX)

14:51:08 <dc_rdfig> Titled item A.

14:52:11 <bitsko> A:Does two things exceedingly well: works on SAX streams (instead of DOM), does only transforms (not a stylesheet language)

14:52:11 <dc_rdfig> Added comment A1.

14:53:12 <bitsko> A:The transformation syntax is based on XSLT, but is a little more procedurally focused since it must work in document (stream) order.

14:53:13 <dc_rdfig> Added comment A2.

14:53:38 <bitsko> A:The relevance to RDF, you ask? scraping made quick and easy.

14:53:39 <dc_rdfig> Added comment A3.

14:55:24 <bitsko> A:Aside, it also appears to fit the bill for a task I've written code for before: transforming XML into application objects.

14:55:25 <dc_rdfig> Added comment A4.

15:30:44 <danbri_> AndyS? are you likely to attend www2003?

16:35:44 * jhendler waves at DanC

16:36:07 * DanC waves

16:37:37 <danbri_>http://www.www2003.org/programme.htm

16:37:37 <dc_rdfig> B: http://www.www2003.org/programme.htm from danbri_

16:37:44 <danbri_> B:|WWW2003 programme

16:37:44 <dc_rdfig> Titled item B.

16:38:06 <danbri_> B:Anyone know if they will support ad-hoc BOFs during the conference?

16:38:06 <dc_rdfig> Added comment B1.

16:38:46 <danbri_> B:I'd like to see the RDF Query discussions continue, but we don't have a formal workshop or suchlike booked for it.

16:38:46 <dc_rdfig> Added comment B2.

16:39:19 <danbri_> B:A room, network, irc logger, whiteboard... probably don't need much more...

16:39:19 <dc_rdfig> Added comment B3.

16:39:21 <jhendler> I've been trying to get a BOF on Sem Web and Topic Maps organized, no one seems to know exactly how BOFs being done

16:39:38 <danbri_> Jim, that's interesting! We should combine our ignorance :)

16:39:45 * danbri_ doesn't know either

16:40:09 <danbri_> RDF<->XTM is something I'm considering for a SWAD-E virtual workshop...

16:40:26 <DanC> virtual? as in no airplanes?

16:40:50 <danbri_> As in... glorified IRC chat with a bit more structure, planning.

16:41:02 <DanC> A:I just found this the other day...

16:41:03 <dc_rdfig> Added comment A5.

16:42:08 <danbri_> A:Is this in the 'schema annotation' space? eg. as way of generating RDF triples when patterns match...?

16:42:08 <dc_rdfig> Added comment A6.

16:42:16 <DanC> A5:yes, this does look nifty. [http://rdfig.xmlhack.com/2003/03/19/2003-03-19.html#1048050245.495870|earlier item 19Mar]

16:42:16 <dc_rdfig> Replaced comment A5.

16:42:59 <DanC> A:May 8 - Oliver Becker presents STX at XML Europe 2003

16:43:00 <dc_rdfig> Added comment A7.

17:02:07 <libby> DanC, when you said (re calendar meet) "I'm inclined to spend the time working on how-to documentation" did you mean rather than being at the meeting or for discussion at the meeting....?

17:02:28 <libby> I like it as an agenda item

17:02:53 <libby> e.g. suggestions for faq

17:04:42 <DanC> I meant it for all the items.

17:04:56 <DanC> i.e. let's spend the whole time working on the wiki.

17:05:27 <libby> sounds good to me.

17:05:41 <libby> - and it's not like people are clamouring to do other stuff ;)

17:05:43 <ronwalf> Hm, does cwm drop the xml:lang attribute on tags?

17:06:22 <dajobe> likely

17:06:38 <ronwalf> ah, ok, not that important

17:06:49 <ronwalf> (to me)

17:07:14 <danbri_> I liked DanC's suggestion, which i took as an agenda proposal

18:29:00 <jhendler>http://www1.internetwire.com/iwire/release_html_b1?release_id=51692

18:29:02 <dc_rdfig> C: http://www1.internetwire.com/iwire/release_html_b1?release_id=51692 from jhendler

18:29:24 <jhendler> C:| Semantic Web Application in the news?

18:29:24 <dc_rdfig> Titled item C.

18:29:55 <jhendler> C: US Democratic National Committee's plans to win back the White House in 2004

18:29:55 <dc_rdfig> Added comment C1.

18:30:02 <jhendler> C: appear to include use of Sem Web

18:30:02 <dc_rdfig> Added comment C2.

18:30:39 <jhendler> C: (March 4th article)

18:30:39 <dc_rdfig> Added comment C3.

18:44:02 <golbeck_zzz> golbeck_zzz is now known as golbeck

18:53:18 <soccos|away> can someone explain the "[[ ... ]]" idiom to me?

18:53:21 <soccos|away> soccos|away is now known as soccos

18:54:00 <DanC> I use [[ to bracket quoted stuff

18:54:06 <DanC> across lines in IRC ]]

18:54:16 <DanC> in case somebody else gets a line in between.

18:55:19 <soccos> and Libby seems to use it to quote lines of RFCs?

18:55:47 <soccos> err, and other things

18:56:08 <deltab> a more-robust form of quote marks

18:56:21 <DanC> so... what's to explain? you seem to understand it thoroughly, soccos.

18:56:25 <libby> I use it for quoting stuff. can;t remmeber where it came from

18:57:33 * soccos is enlightened

18:58:12 <soccos> I was just looking at your tz fixes libby

18:58:19 <libby> I mean, I can't remember where the convention came from....

18:58:23 <libby> oh yes?

18:58:43 <libby> any comments?

18:58:48 <soccos> I've noticed some inconsistencies between iCal and evolution too

18:59:18 <soccos> evolution didn't like an iCal file I fed it

18:59:39 <soccos> appeared to be TZ realated but unconfirmed

18:59:42 <libby> really? did you work out why?

18:59:48 <soccos> :(

19:01:06 <libby> could maybe see if it was the tz bit - fancy trying http://www.w3.org/2002/12/cal/test/20030115mtg.ics? and then editing the ics file to remove the tz portion and trying again?

19:01:25 <soccos> the various RFC2445 implementations seem to have some incompatibilities

19:01:36 <libby> it's very complex...

19:02:04 <soccos> where did the above .ics come from?

19:02:25 <libby> it started life as an apple ical file. I just added timezone stuff to it

19:02:51 <libby> actually this one might be better: http://www.w3.org/2002/12/cal/test/20030122mtg.ics

19:02:52 <soccos> you want me to open it in evolution?

19:03:22 <soccos> I'm more of a mozilla calendar user today

19:06:51 <soccos> the events in that calendar don't appear to show up in mozcal

19:10:23 <jhendler> sigh - Ian and Sean have sent email to the group - they both say they prefer the sharing stuff - so I'm really confused now...

19:10:36 <jhendler> [[bleed from another chat - sorry]]

19:10:56 <libby> soccos, only if you want to :)

19:11:10 <soccos> I just did, doesn't show up in mozcal

19:11:15 <libby> interested in intereop issues that's all. not used evolution much

19:11:25 <soccos> should be in January 2003 right?

19:11:29 <libby> bah, that's bad news

19:11:32 <libby> yep

19:12:10 <soccos> want me to hack the TZ?

19:12:44 <libby> why do you think it doesnt work?

19:13:19 <soccos> dunno yet, I actually did an import which discarded all of the tz fields anyway

19:13:27 <libby> ah.

19:13:40 <libby> well I'd be interested if you feel liek investigating

19:13:42 <soccos> as mozcal isn't very good at opening existing .ics files

19:13:48 * soccos investigates

19:14:11 * libby thinks probably a tz problem - wont recognise the timezone ids maybe?

19:14:35 * libby wonders if she created monster icalendar files that wont work anywhere at all...

19:14:36 <soccos> no, looks like a VALARM issue

19:16:03 <libby> oh.

19:16:12 <libby> hm, thought I had that workign before

19:16:21 <libby> between apple and moz

19:21:14 <soccos> huh, weird intermittent behaviour

19:22:15 <soccos> it was fine when I imported this time

19:23:06 * DanC is in a meeting, kinda bummed to be missing tz discussion

19:25:15 <libby> erk

19:25:33 <libby> moz calendar devleopment is being scaled back :(

19:27:33 <soccos> really?

19:29:05 <libby> that's odd, thought I sent the message to www-rdf-calendar

19:29:34 <libby> shame :(

19:29:55 <libby> it's all voluntary, but the paid-for project it overlapped with has been scaled back

19:30:16 <soccos> @ oeone?

19:30:37 <libby> yep

19:30:41 <soccos> btw, your calendar seems to work in mozcal

19:30:57 <libby> hooray

19:31:08 <soccos> I just got some coaching from mikep in opening existing .ics files in mozcal

19:31:38 <soccos> despite numerous messages suggesting it's going to overwrite your file... it doesn't ;-)

19:32:15 <libby> heh

19:32:46 <soccos> there was just one vevent right?

19:32:57 <soccos> anything else you want me to check in mozcal?

19:33:00 <libby> yep

19:33:18 <libby> well, any ics you fancy from that test dir would be great

19:33:29 <libby> there's only one there originally from moz I think

19:34:03 <soccos> I think I've run out of time for this

19:34:09 <libby> no prob

19:34:17 <libby> they're pretty similar anyway

19:34:17 <soccos> but I can do more testing if you need

19:34:58 <soccos> I'm intrigued to see if the alarm goes off now

19:34:59 <libby> it would be good to know if those files interoperate, because I edited them. If you do do any, would you mail to the www-rdf-calendar list or use the wiki?

19:35:02 <libby> heh

19:35:43 <soccos> what were we testing there?

19:35:56 <libby> good question

19:36:16 <libby> we need some ics files to start with to generate thr rdf files

19:36:29 <libby> and we had some...but not quite right a/c rfc 2445

19:36:41 <libby> here's that mail from Mike btw about moz cal: http://groups.google.com/groups?dq=&hl=en&lr=&ie=UTF-8&threadm=b5n37a%24b361%40ripley.netscape.com&prev=/groups%3Fq%3Dnetscape.public.mozilla.calendar%26ie%3DISO-8859-1%26hl%3Den%26btnG%3DGoogle%2BSearch

19:36:48 <libby> erk, nasty url

19:37:18 <libby> really we have to go back to the rfc, and that's what I've tried to do. but interoperating would be a bonus :)

19:37:21 <soccos> make a shorter link ;-)

19:45:11 <soccos> soccos is now known as soccos|away

19:55:43 <soccos|away> soccos|away is now known as soccos

19:56:42 <ericP> hey libby, i wrote a prog to look for functional dependencies between RDF relations in a triple store

19:56:48 <ericP> it's pretty keen!

19:57:17 <ericP> it currently proposes a bunch of tables to store the most common data

19:59:41 <libby> cool!

19:59:49 <libby> got a link?

20:00:30 <ericP>http://dev.w3.org/cvsweb/perl/modules/W3C/Rdf/bin/compileTables

20:00:31 <dc_rdfig> D: http://dev.w3.org/cvsweb/perl/modules/W3C/Rdf/bin/compileTables from ericP

20:01:14 <ericP> D: perl and SQL tool for compiling triples from a generic triple store to specialized tables.

20:01:14 <dc_rdfig> Added comment D1.

20:01:54 <ericP> it's fallen a bit into disuse in the last 88 minutes

20:02:31 <libby> D:neat :))

20:02:31 <dc_rdfig> Added comment D2.

20:02:42 <libby> ericp, got any writeup or anything?

20:03:24 <ericP> grep \# compileTables

20:03:28 <ericP> nope, not really

20:03:38 <ericP> s/really/at all/

20:04:06 <ericP> part of the code takes a random subset of the data for quicker analysis

20:04:36 <libby> why don;t you describe what it does here, and then you can link to the logs :)

20:04:54 <ericP> makes sense

20:05:08 <libby> "I'm listening"

20:05:23 <ericP> i'm taking a deep breath

20:05:33 <libby> :)

20:06:12 <danbri_> ericP, so its data driven, looks for likely functional and inverse-functional properties, based on patterns in the currently-loaded data?

20:06:41 <ericP> not sure about inverse-functional

20:07:04 <ericP> first you sample a subset of the triple store

20:07:38 <ericP> start with a target number of statements and a maximum number of statements that can come from a single document

20:08:12 <ericP> pick a random statement and get it's document number (attribution)

20:08:56 <ericP> if the document has more than maximum statements, pick a new random statement

20:09:23 <ericP> copy all the statements from that document (attribution) to a working table

20:10:14 <ericP> if the working table is still smaller than the desired subset, pick a new statement

20:10:22 <soccos> soccos is now known as soccos|away

20:10:30 <ericP> now we have a little statements table

20:10:44 <ericP> SELECT predicate,COUNT(*) FROM $self->{OPERTABLE} GROUP BY predicate

20:11:00 <ericP> this finds the most common predicates

20:11:27 <ericP> starting with the most common predicate:

20:11:44 <ericP> SELECT ap.predicate,count(*) FROM $self->{OPERTABLE} AS a INNER JOIN $self->{OPERTABLE} AS ap ON a.subject=ap.subject WHERE a.predicate=$p AND ap.predicate!=$p GROUP BY ap.predicate

20:12:13 <ericP> this finds the predicates that are associated with the predicate you are investigating

20:12:31 <ericP> starting with the most common associated predicate:

20:12:50 <ericP> SELECT count(*) FROM $self->{OPERTABLE} AS ap LEFT OUTER JOIN $self->{OPERTABLE} AS p ON ap.subject=p.subject WHERE ap.predicate=$ap AND p.predicate=$p AND p.predicate IS NULL

20:13:08 * libby following so far

20:13:45 <ericP> test to see if the associate property ever comes from a subject which the original (candidate) predicate does not

20:13:58 <ericP> # Find count(*) of '((?ap ?s ?o2)!(?p ?s ?o))

20:14:10 <ericP> that's the algae-related comment

20:14:17 <ericP> example:

20:14:43 <ericP> anot1 annotates http://www.w3.org/

20:14:58 <ericP> anot1 rdf:type Comment

20:15:26 <ericP> this would make it seem that you should store type in the same table as the annotates property

20:16:25 <ericP> but plenty of rdf:type arcs come from subjects that don't have an annotates property so it can't go in the same table (interesting reasons for this; take it on faith for now)

20:17:13 <ericP> so now we've eliminated arcs that apply to things other than the domain of the candidate predicate

20:17:57 <ericP> the error messages for this is " -- loose associations\n" to imply the associated property is some sort of floozy with loose more character

20:18:16 <ericP> # Find count(*) of '((?ap ?s ?o2)(?ap ?s ?o3))

20:18:32 <ericP> this means we are looking for repeated properties.

20:19:08 <ericP> since there is no way to store them in the database and query them again (as SqlDB currently compiles SQL queries), these are eliminated now too

20:19:24 <ericP> SELECT COUNT(*) FROM $self->{OPERTABLE} AS ap INNER JOIN $self->{OPERTABLE} AS rp ON rp.subject=ap.subject WHERE ap.predicate=$ap AND rp.predicate=$ap AND rp.id!=ap.id

20:19:30 <ericP> is the query to do this

20:20:04 <ericP> finally, see how often the associated property is used with the candidate property

20:20:33 <ericP> query actually see how many times it is not use with the candidate property

20:20:35 <ericP> SELECT COUNT(*) FROM $self->{OPERTABLE} AS a LEFT OUTER JOIN $self->{OPERTABLE} AS ap ON a.subject=ap.subject WHERE a.predicate=$p AND ap.predicate=$ap AND ap.predicate IS NULL

20:22:15 <libby> cool :)

20:22:25 <libby> ericp, what made you do the sampling?

20:22:39 <ericP> there's a minimum association ratio (by preference). if it's below the ratio, pick another staetment for canidate property

20:22:54 <ericP> in the end, you have a pair of statements.

20:23:40 <ericP> when you tell the library about it, it looks to see if it had any other properties associated with the candiate property and adds the associate property to the list

20:24:20 <ericP> reporting currently just spits out the reccomendations. doesn't do any creating tables or moving data out of the triple store

20:24:25 <ericP> .

20:24:37 <ericP> sampling: it was too slow. it was pissing me off.

20:24:44 <ericP> i got cranky

20:24:46 <libby> heh

20:25:00 <libby> but it would work, just slowly, without sampling

20:25:07 <ericP> yup

20:25:26 <ericP> it gets slower really quickly with more statements in the db.

20:26:11 <ericP> sampling code is kinda cool, though. you can use it to make a copy table for other expensive heuristic opperations

20:26:40 <libby> it does sound cool

20:26:53 <libby> so, how dependent is the code on other code?

20:27:03 <libby> particular databases (mysql?)

20:27:07 <ericP> not much. needs the database interface

20:27:13 <ericP> mysql, probably pretty bad

20:27:26 <ericP> `show create table Statements`

20:27:32 <ericP> don't know that anyone else does that.

20:27:32 <libby> recently I've been cutting my code right back to simple bits of usefulness

20:27:36 <ericP> the idea was you don't want to look

20:27:52 <ericP> at all the data, but you want a good representation

20:28:01 <libby> sure

20:28:14 <ericP> a good representation would entail taking entire, modestly sized documents.

20:28:54 <libby> what about other code dependencies.?

20:28:55 <ericP> cutting back, yeah, i had the idea that this wouldn't be terribly portable

20:29:10 <libby> somethign standalone would be very useful

20:29:17 <libby> yeah, sorry :)

20:29:17 <ericP> oh, for the sampler?

20:29:29 <libby> I was thinking whole thing, but yeah....

20:30:09 <libby> I'd liek to run over my code but I have postgres and noone the w3c stuff installed at the moemnt

20:30:22 * libby not being critical, just think sounds really useful...

20:30:35 <ericP> sampler takes a name of a table to duplicate, its primary key, a minimum sample size, and a maximum size for an individual contribution

20:30:50 * ericP cries

20:31:05 * ericP ... and takes it personally

20:31:10 <libby> aw

20:31:56 <ericP> i think it needs certain functionality in the triple store.

20:32:13 <ericP> could make names of fields variables as is the promary key currently

20:32:56 <libby> you mean the sql db schema for triples? what does your look like again?

20:32:59 <ericP> but it would require that some field how the document name (attribution (provenience (subgraph ID)...

20:33:16 <ericP> s/how/know/

20:33:21 <libby> I think most/many have that

20:33:27 <ericP> yeah, me too

20:33:32 <ericP> Statements:

20:34:43 <ericP> id (primary key) | predicate (int) | subject (int) | object (int) | reificationId (int or NULL) | reificationBag (int or NULL) | attributions (int)

20:34:53 <ericP> all the above ints are really foreign keys

20:35:21 <libby> right

20:35:53 <libby> I dont do reification spcially, but the rest is the same I think. I just bung in the url opf teh doc rather than an int for it.

20:35:55 <ericP> RdfIds: id (pk) | type enum('Ref','String','Gen','Table') | genId | uri | string | tableName | tableRwoId

20:36:41 <ericP> uri points to Uris table

20:36:51 <ericP> string points to String table (with MD5)

20:36:51 <libby> what does the rdfids do?

20:37:26 <ericP> intermediarry bewteen Staetments.{predicate,subject,object...} and the actual data tables for the types of nodes

20:37:50 <libby> ah, ok

20:38:01 <ericP> i think i can re-do this wihtout a separate Uris,Strings,GenIds table

20:38:17 <ericP> it would cost a little bit in calculating MD5s

20:38:23 <libby> got that info somewhere, maybe in triples, maybe in the strings/uris table that triples points to.

20:38:28 <libby> (I have I mean)

20:38:40 <ericP> need a nice short and rich index candiate.

20:39:42 <ericP> tableName and tableRowId are for linking to entities stored in other tables like the ones i'm trying to generate with compileTables

20:39:50 <libby> right

20:40:25 <ericP> hey, do you know much about Versa?

20:40:33 <libby> not a lot, no

20:40:42 <ericP> i'm tyring to figure out if Uche's examples are protocol or API

20:41:06 <libby> I was mailing uche when I was wroking with your document for my converter

20:41:11 * libby looks

20:41:13 <ericP> also, could you give me a pointer to this chat so i can link to it from compileTables?

20:41:21 <libby> hm

20:41:27 <libby> logger, pointer?

20:41:34 <libby> heh, perhaps not

20:41:35 <ericP> uh oh

20:41:47 <libby> ah, logger_1

20:41:52 <libby> logger_1, pointer?

20:41:52 <libby> See http://ilrt.org/discovery/chatlogs/rdfig/2003-03-25#T20-41-52

20:41:56 <ericP> tx

20:41:58 <libby> hooray!

20:44:29 <libby> D:|Eric Prud'hommeaux's new tool for generating specific tables from a generic triple store

20:44:29 <dc_rdfig> Titled item D.

20:45:28 <libby> D:see [http://ilrt.org/discovery/chatlogs/rdfig/2003-03-25.html#T20-00-30|EricP documenting it live on #rdfig :)]

20:45:28 <dc_rdfig> Added comment D3.

20:46:51 * libby gathers together 70-odd icalendar testcases from various places...

20:48:08 <ericP> libby, i have a more specific pointer: http://ilrt.org/discovery/chatlogs/rdfig/2003-03-25#T19-56-42

20:48:20 <ericP> but i don't know how to update comments

20:48:27 <libby> ok

20:48:31 <libby> do this

20:49:08 <libby> D3:see [http://ilrt.org/discovery/chatlogs/rdfig/2003-03-25#T19-56-42|EricP documenting it live on #rdfig :)]

20:49:09 <dc_rdfig> Replaced comment D3.

20:49:14 <libby> that what you meant?

20:49:41 <ericP> indeed - tx!

20:49:59 <libby> cool

20:52:06 <ericP> how do i query comments on a URI?

20:52:33 <danbri_> ericP, em at one point had this stuff piped into Annotea

20:52:52 <ericP> cool

20:53:04 <danbri_>http://www.w3.org/2001/09/chump/

20:53:04 <dc_rdfig> E: http://www.w3.org/2001/09/chump/ from danbri_

20:53:10 <danbri_> E:|annoChump Overview

20:53:10 <dc_rdfig> Titled item E.

20:53:15 <ericP> 'course it would probably have no idea what anupdate was

20:53:27 <ericP> replaced data right next to regular data

20:53:48 <danbri_> does that page help you find the data in annotea/algae'able form?

20:54:28 <ericP> not sure i get you

20:55:00 <ericP> is the question "does it link to the algae HMTL interface?"

20:55:01 <ericP> ?

20:55:51 <ericP> BTW, i ran the compiler on annotest. it proposed 6 tables, five made sense to me

20:56:20 <libby> that's good going

20:57:09 <danbri_> what was the table that didn't make sense?

20:57:29 <ericP> tables for annotation nodes, html message (associated wiht the annotation) nodes, earl assertor nodes and one wierd product of earl reifications: http://www.w3.org/1999/02/22-rdf-syntax-ns#subject nodes

20:57:36 <ericP> the last one

20:57:49 <danbri_> question was: I found this old page about annotea harvesting of #rdfig chump metadata... does that answer your "how do i query comments on a URI?" question?

20:58:02 <danbri_> yuck reification

20:58:03 <ericP> ahh, got it.

20:58:51 <ericP> i was actually interested in the data that logger had

20:59:04 <dajobe> ah

20:59:07 * dajobe has a plan

20:59:51 <ericP> ie, did libbies first (D:see...) assertion create a node for #T20-00-30 that didn't get replaced by #T19-56-42?

21:00:07 <dajobe> the chump isn't rdf

21:00:10 <dajobe> it stores things as xml

21:00:17 <dajobe> then uses xslt to transform them

21:01:28 <dajobe> see http://rdfig.xmlhack.com/index.xml

21:41:08 <danbri_>http://www.w3.org/2000/10/swap/setup.py

21:41:09 <dc_rdfig> F: http://www.w3.org/2000/10/swap/setup.py from danbri_

21:41:21 <danbri_> F:|A not-working-yet attempt at a setup.py installer for Cwm/SWAP

21:41:21 <dc_rdfig> Titled item F.

21:41:52 <danbri_> F:See also [http://dev.w3.org/cvsweb/2000/10/swap/|SWAP filetree] in dev.w3.org (lags main site, by ?1hr?)

21:41:52 <dc_rdfig> Added comment F1.

21:42:24 <danbri_> F:Clues, suggestions welcomed, here or in [http://esw.w3.org/t/view/ESW/CwmTips|CwmTips] Wiki

21:42:24 <dc_rdfig> Added comment F2.

21:42:45 <ericP> clues about why dev lags?

21:43:02 <ericP> it's replicated from www.w3.org via cron'd rsync

21:43:05 <danbri_> F2:Any clues, suggestions on making setup.py work are welcomed, here or in [http://esw.w3.org/t/view/ESW/CwmTips|CwmTips] Wiki

21:43:05 <dc_rdfig> Replaced comment F2.

21:43:08 * timbl can't find type= attr in new RDF syntax doc

21:43:10 <danbri_> sorry :)

21:43:36 <danbri_> type was special cased in old syntax?

21:43:55 <timbl> Has that gone now?

21:44:00 * danbri_ looks

21:44:01 <timbl> I was chasing test case errors

21:44:08 <ericP> only in the sense that a typedNode implied a type arc

21:44:22 <danbri_>http://www.w3.org/TR/1999/REC-rdf-syntax-19990222/#grammar

21:44:22 <dc_rdfig> G: http://www.w3.org/TR/1999/REC-rdf-syntax-19990222/#grammar from danbri_

21:44:24 <timbl> type= was werid beccause it took a URI not a string - made a symbol not a lit

21:44:25 <ericP> (re type was special cased)

21:44:36 <danbri_> G:|Old Model and Syntax REC, 'formal grammar'

21:44:36 <dc_rdfig> Titled item G.

21:44:49 <danbri_> G:[6.11] typeAttr ::= ' type="' URI-reference '"'

21:44:49 <dc_rdfig> Added comment G1.

21:44:59 * timbl checks the test case

21:45:27 <danbri_> y'know, I never noticed that...

21:46:43 <timbl> ######### from positive parser test rdft_test0006.nt: cwm http://www.w3.org/2000/10/rdf-tests/rdfcore/rdf-ns-prefix-confusion/test0006.rdf

21:47:13 <timbl> ref: -<http://example.org/resource/> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <http://example.org/class/> .

21:47:26 <timbl> c wm: <http://example.org/resource/> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> "http://example.org/class/"

21:47:30 <danbri_> G:No mention in [http://www.w3.org/TR/rdf-syntax-grammar/#section-Syntax-property-attributes|property attributes] bit of new syntax spec. Am I looking in wrong place?

21:47:30 <dc_rdfig> Added comment G2.

21:47:39 <ericP> how is that different from [6.7] aboutAttr or [6.18] resourceAttr

21:47:42 <ericP> ?

21:48:12 <danbri_> implies a type arc, whose value is the resource named in the attribute's contents

21:48:14 <danbri_> i think

21:48:24 <timbl> tyis was propertyAttr

21:48:25 <danbri_> hmm does it take relative uris?

21:49:02 <taazzzz> taazzzz is now known as taaz

21:51:05 <dajobe> type's a special case in the grammar *action* for property attr/element

21:51:33 <dajobe> G:type's a special case in the grammar *action* for property attr/element. Look in the grammar.

21:51:33 <dc_rdfig> Added comment G3.

21:52:07 <danbri_> G:Thanks Dave :)

21:52:07 <dc_rdfig> Added comment G4.

21:55:00 * timbl action ....look in teh grammer ... "type" not found in the table

21:55:34 <dajobe> timbl: no, it's in the action for the propertyElt

21:55:51 <dajobe> the table is just a summary of the rule matches

21:56:02 <timbl> And the actions are in section 7.

21:56:16 <dajobe> yes

21:57:52 * timbl sees 7.2.14 Production propertyElt

21:58:08 <timbl> Has special case for rdf:li

21:58:50 <dajobe> that's earlier

21:59:07 <dajobe> but you can just do it whenever you like

21:59:37 <timbl> I see... 7.2.11 Production nodeElement step S2.

21:59:38 <dajobe> it's recognised when an element or attribute of that name is found, then like a macro, the right rdf:_n name is made

22:00:03 <timbl> It was type I was looking for.

22:01:31 <timbl> Ok, thanks.

22:17:27 <timbl> A lot of changes from 1.0.

22:17:55 <timbl> I wonder why, when making lists, the statemenmt htat something is a list is added?

22:18:33 <DanC> I think cuz that's the way it was originally specified/implemented in DAML, and Mike Dean came to rely on it, and it sorta stuck.

22:18:50 <DanC> i.e. you can do each(rdf.type, rdf.List)

22:18:52 <timbl> List is the domain of rdf:first - if you put one you really don't need the oethr.

22:19:20 <timbl> each?

22:19:22 <DanC> domain isn't known by RDF parsers.

22:19:30 <DanC> each in the sense of the swap API

22:19:30 <timbl> ok, api each

22:19:54 <dajobe> yes, parsers need not know any schema terms except for the few in syntax - and the special cases - gah!!

22:19:55 <timbl> In a daml system you can still do each(type, List) without explicit statements.

22:20:13 <DanC> schema terms in syntax?!?!??!

22:20:16 <dajobe> I only got it (daml collections) when I wrote it out like http://www.w3.org/TR/rdf-primer/#collections

22:20:20 <dajobe> rdf:type for example

22:20:33 <timbl> The fact that domain isn't known by parser is no excuse for them outputting trivia for the sake of it.

22:21:08 <DanC> rdf:type ... oh... I thought by "schema terms" you meant stuff in the rdfs namespace

22:21:12 <dajobe> no

22:21:33 <dajobe> I meant vocab beyond syntax architecture (rdf:Description, rdf:li and so on)

22:22:26 <DanC> first/rest was something of an iffy thing for RDF Core to do, charter-wise... it was presented as a mature, worked-out design. And the way the design worked was to spit out List tripes along with first/rest.

22:22:46 <timbl> Not in my parsers

22:22:55 <DanC> so by the time the RDFCore WG got around to examining the issue, folks relied on the List triples.

22:23:02 <timbl> I've been using collections all this time and never spat them out

22:23:18 <timbl> relied on -- crud.

22:23:32 <timbl> Why oh why can't we build something minimal?

22:24:03 <timbl> How about outputting { x a daml:Thing} for each node just so we know?

22:24:08 <jhendler> priblem is we only needed the damned collections to contain a list with an end, I had hoped RDF Core would develop something simpler and WOWG would move there -- didn't happen

22:24:24 <timbl> I am not complaining about list.

22:24:29 <DanC> because minimalism and consensus compete. and our process very explicitly values consensus, and only implicitly and secondly values minimalism.

22:24:53 <timbl> I think the existence of lists had to happen somewhere and a high layer would have been cool but hey lets put it in rdf.

22:25:15 <timbl> But that is a different thing from sticking unnecessary triples (lots) in everyone's data store.

22:25:27 * DanC finds it pretty frustrating that timbl doesn't really think about WG charters until this point in the game

22:25:28 <jhendler> maybe, but the first, rest, li stuff seems -- yes, just what I was going to say

22:25:56 <DanC> unnecessary: says you.

22:26:23 * timbl finds it frustrating that "oh, and make it a good design" has to be written into charters.

22:26:33 <DanC> good according to whom?

22:26:39 <timbl> Voila.

22:26:47 <timbl> That's why you can't write it into the charter.

22:26:48 * jhendler would be curious how one would operationalize "make it minimal"

22:27:10 <timbl> I suppose I am sighing that the minimal design idea wasn't mroe closely shared.

22:27:23 <timbl> Its like porn. I knwo it when I see it.

22:27:32 <timbl> (unnecesary)

22:27:37 <DanC> perhaps I should be more clear: RDF Core wasn't chartered to do design *at all*.

22:28:01 <jhendler> but I do think consensus does fight w/minimality (i.e. the "compromises" don't tend to favor the simplest solution, but the ones with "someting for everyone")

22:28:09 <timbl> unnecessary in that there is a well-defined algorthim in linear time for regurgitating them -- they provdie no extra data.

22:28:48 <timbl> At this layer in the architecture, we an't afford croft.

22:29:25 <DanC> "At this layer in the architecture, we an't afford croft." <- *that's* what you could have told the RDF Core WG a year or so ago but did not.

22:29:51 <DanC> you made this bed, and you will please now lie in it.

22:30:48 <DanC> hmm... let me remove some heat...

22:31:22 <DanC> When we made one of the first backwards-incompatible changes... perhaps aboutEach or aboutEachPrefix... there was an action to ask the SemWeb CG if that was in scope of the charter.

22:31:32 <DanC> I tried to find the outcome of that action, but I think it got dropped.

22:32:03 <DanC> It could have been very healthy to have some discussions in the CG (with timbl perhaps invited) about the balance between the past and the future.

22:32:52 <jhendler> +1

22:33:01 <DanC> I tried to get some management discussion of the whole datatypes mess too... not as hard as I could have, I suppose... but again, we spent a long time coming to a decision that we're not sure anybody's happy with.

22:33:56 <DanC> now last call is perhaps not a completely insane point to have these discussions... it's better than doing it after REC, I suppose...

22:34:02 * jhendler ws-chor telecon over,back to real work - g'night.

22:34:21 <DanC> ... but it's pretty close to insane.

22:34:49 <DanC> i.e. it's not very respectful of the engineering investment to date... the attempts of the WG members and chairs to execute the charter they were given.

22:35:36 <timbl> I clearly should have put cwm though the parser tests before. I simply had not dreamed that one was supposed to add those unnecessary triples.

22:36:04 <DanC> unnecessary to you.

22:36:27 <timbl> I do note that this is *no* concerned with removing stuff from teh old 1.0 spec)

22:37:01 <timbl> unnecessary to anyone in that they contain zero extra information. 0.0000000

22:37:30 <timbl> You can take any graph without them and generate the graph with them.

22:37:37 <DanC> at non-zero cost.

22:38:01 <timbl> But with no extra information. So they are redundant.

22:38:23 <DanC> yes, they are redundant.

22:38:29 <timbl> Cost is finding every x first y and adding x type list.

22:40:40 <danbri_> do we say domain of 'first' is List?

22:40:55 * danbri_ thinks a smart parser should be able to supress such things

22:41:11 <DanC> I could be more clear on another thing: I'm not invested in the first/List stuff, but I'm concerned about a process where the team is represented in the decision-making, and then at last call, the WG gets input from the team from all over the map.

22:42:15 <DanC> perhaps the List tests are information that you couldn't have predicted at the time...

22:42:26 <DanC> but reification semantics is another story.

22:43:56 <DanC> well, perhaps the team isn't the only problem with the reification decision-making... the chair had an action to make sure fix-or-nix-reification was in the issues list and discharged it by pointing to some issue about statement equality.

22:44:18 <DanC> so I was sorta expecting a particular issue I could bring to SWAD, but it never materialized.

22:45:16 <timbl> The only project I've found which actually used reification was CORC and they used it as though it were quoting.

22:45:51 * DanC doesn't seem to be getting thru

22:46:19 <DanC> tim, the WG had the "who's using it anyway?" discussion. And the de-re/de-dicto discussion. Nothing you're saying is new.

22:46:24 <timbl> Yes, danbri, a smart parser can suppress error on use of reification and suppress redundant List triples.

22:47:29 <DanC> What if MemberCo XYZ participated in the resolution of some issue. They were party to the decision. Then they asked the WG to do their work over again without presenting any new information. We'd be pretty flargin annoyed, no?

22:47:35 <timbl> You say nothing I am saying is new, but I said "The only project I've found which actually used reification was CORC and they used it as though it were quoting." and no one has said that.

22:48:00 <timbl> people have said that reification is needed with de-re semantics and no quoting by a user community .

22:48:19 <DanC> yup

22:49:01 <timbl> The new information is that if you actually ASK eric m about how it was used, he says taht the de-re interpretation's effect "would have been regarded as a serious error"

22:49:09 <DanC> that's the way the deployed code works, and the users are using it, and their representatives in the WG weren't willing to have it removed.

22:49:28 <DanC> We did ask Eric M. This is *not* new information.

22:49:42 <timbl> People are not usingthe deplyed parser with the deployed infrence engines. This heading for a crash.

22:49:53 <timbl> Actually very few people seem to be using reification at all.

22:50:24 <timbl> So you found that the WG's appraoch was not shared by the user community?

22:50:35 <DanC> "very few" in your estimation. Sufficiently many, in the view of WG participants, that W3C owes them continued support of REC-level-endorsed functionality.

22:50:53 <timbl> It's not new inform?åiBBBhjhj\GqfdwdfuIpr

22:51:23 * timbl gets upset -> types fast -> hits alt-space -> switches o jp entry mode by mistake

22:51:56 <DanC> you seem to want to charter a WG, i.e. delegate decisions, except when you don't like the outcome.

22:52:07 <timbl> :)

22:52:26 <em> (about to head off) but notes other projects i'm aware (including CORC) of that use reification is Carmen and GEM ... there are others (but again not sure how they're specifically using this). CORC was the only one I was 'familier' with because I helped in the arch/design

22:53:05 <DanC> The point about "not usingthe deplyed parser with the deployed infrence engines" is perhaps well-made, but it goes to the charter discussions again!

22:53:10 <timbl> So, em, the WG asked you, and you said that the way CORC used it would have been screwed up by de-re semantics?

22:53:32 <timbl> Did you tell the WG what you told me?

22:53:48 <DanC> CORC wasn't the only application the WG considered.

22:54:17 <em> yes to timbl, and agree with danc - i recall several others were mentioned

22:54:31 <timbl> Its the only one where I have got to the source. And I was lead to eblive it supported the WG, but it doesn't.

22:54:34 <DanC> and I don't think EricM learned about CORC being screwed up after we made the decision. This information was available to the WG when the decision was made.

22:55:26 <timbl> My impression (perhaps wrong) is that the implications of the de-re semantics had never occured to the CORC project.

22:55:38 <DanC> perhaps looking forward is better: next time, let's be sure we go on record as objecting, rather than abstaining or supporting decisions that we're lukewarm about on grounds of minimalism etc.

22:55:43 <em> and since CORC was being migrated into Connexion (and this whole context-sensitive display build by reifaction) seemed to be removed didnt seem as much of a hit

22:56:03 <em> de-re semantics had never occured to the CORC project - correct ; full inference was not implemented

22:56:40 <DanC> bnodes almost didn't make it into RDF semantics because the WG wasn't chartered to look at query stuff. Only on the grounds that entailment tests were interesting did they go in.

22:57:25 * em has to run; not sure obscure references to other projects helped...

22:57:28 <DanC> our position on xml:lang is hosed, but all the arguments for saying so rely on future work, i.e. query/rules/etc. stuff that, by charter, doesn't count much.

22:58:05 <timbl> So "no new info" does not hold.

22:58:20 <timbl> s/so//

22:58:50 <DanC> ok, so CORC didn't think about de-re/de-dicto. Are we going to keep 20-odd engineers in WG-purgatory for another 3 months over that?

22:59:08 <DanC> If so, I think you owe them a charter change.

22:59:20 <DanC> s/you/we/

22:59:50 <timbl> Not because CORC didn't think so.

23:00:16 <timbl> Not because when they did (a few days ago) they decided the de-re was a serious error.

23:00:40 <timbl> Not because no one else has thought about it of the people who are using it.

23:00:51 <timbl> Not because a minority of applications use it.

23:00:51 <DanC> The WG was *well aware* of the de-re/de-dicto issues. there's an extensive "use/mention and reification" thread in the WG proceedings. This *is not news*.

23:01:35 <timbl> I've read a lot of it.

23:01:37 <DanC> "no one else has thought about it of the people who are using it." <- not so. (you're jumping to conclusions)

23:01:55 <danbri_> it isn't news, but the issue is sufficiently subtle that imho folk are only lately realising how un-useful rdf:pred/sub/obj are

23:01:59 <DanC> the WG has indulged in going over this issue, even though I think your request to do so is out of order:

23:02:35 <DanC> "It isn't the only one, it is just one I knew of, there were hundreds

23:02:35 <DanC> (I didn't bother to count) more references on google." -- Beckett http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Mar/0147.html

23:02:38 <timbl> "not so" What evidence?

23:03:27 * dajobe is lurking

23:04:23 <danbri_> danbri@fireball:~/scutter/webcache$ grep bagID *

23:04:24 <danbri_> danbri@fireball:~/scutter/webcache$ ls *rdf |wc

23:04:24 <danbri_> 493 493 9387

23:04:30 <danbri_> my rdf harvester didn't find any

23:04:31 <DanC> ok, so folks are only lately realizing the impact of de-re/de-dicto... how long do we wait for them? How can an engineer go to his manager with a W3C charter and say "I want to do that" when we yank them around like this at the end?

23:04:35 <danbri_> or is it BagID?

23:04:45 * danbri_ tries caseless; no hits

23:05:05 <danbri_> fair point

23:05:46 * dajobe checks his pile-of-rdf file

23:05:49 <timbl> It also unfair to let a WG go into CR when you don't expect them to get out, because this realization will become greater during that period.

23:05:53 <danbri_> So is there at least one solid use of rdf:predicate/subject/object that isn't broken according to timbl's concerns?

23:06:22 <DanC> into CR? RDF core is headed for PR. There are piles of implementations that pass all the tests. including the reification tests.

23:06:41 <danbri_>http://www.w3.org/TR/EARL10/

23:06:42 <dc_rdfig> H: http://www.w3.org/TR/EARL10/ from danbri_

23:06:57 <danbri_> H:|Evaluation and Report Language (EARL) 1.0, W3C Working Draft 06 December 2002

23:06:57 <dc_rdfig> Titled item H.

23:07:46 <timbl> I guess we have to charter RDF light

23:08:02 <danbri_> H:Used to use rdf reification vocab. Don't think it does now.

23:08:03 <dc_rdfig> Added comment H1.

23:08:10 <dajobe> H:"1.0 uses properties instead of reification"

23:08:11 <dc_rdfig> Added comment H2.

23:08:19 <danbri_> H:Hurrah!

23:08:19 <dc_rdfig> Added comment H3.

23:08:38 * DanC got timbl to comment on the charter issues, whew!

23:08:38 <danbri_> H:I was arguing against reification at the Bristol f2f, forgot what happened afterwards...

23:08:39 <dc_rdfig> Added comment H4.

23:09:01 <DanC> yes, we owe the WG a serious "now... just what did we ask you to do in the first place? is that still what we want?" chat.

23:10:55 <timbl> I know that it is difficult to charter a WG right and difficult for a WG to operate under a charter. But when the same goals and ideas are shared by everyone involved, and we end up making something non-optimal, -- it is worth talking about it

23:12:34 <danbri_> the WG and the Team aren't the only stakeholders... other WGs (CC/PP), w3c members, organisations (DCMI), disorganisations(RSS,...) are all waiting to use the new RECs

23:12:50 <danbri_> is it worth talking about it to all of them?

23:12:53 * DanC answers an invitation to #validator for a bit...

23:13:15 <danbri_> might take less time to just propose some 'health warning' text to accompany rdf:predicate et al

23:13:35 <bitsko> heh, disorganizations ;)

23:17:37 <DanC> the most straightforward way to talk with all those parties, danbri, is to call for review of a Proposed Rec, no?

23:17:49 <DanC> or perhaps talk with them in CR.

23:24:12 <DanC>http://www.w3.org/2001/tag/ilist#namespaceDocument-8

23:24:12 <dc_rdfig> I: http://www.w3.org/2001/tag/ilist#namespaceDocument-8 from DanC

23:24:30 <DanC> I:namespaceDocument-8 : What should a "namespace document" look like?

23:24:30 <dc_rdfig> Added comment I1.

23:24:37 <DanC> I:1 year old today!

23:24:40 <dc_rdfig> Added comment I2.

23:24:46 <dajobe> I:if you could decide that by friday, it would be nice

23:24:46 <dc_rdfig> Added comment I3.

23:25:04 <DanC> I:of course, the issue really goes back to '99 and before.

23:25:04 <dc_rdfig> Added comment I4.

23:37:02 <DanC> "no one else has thought about it of the people who are using it." <- I read that too fast... I read it as "no one else has thought about the people who are using it."

23:37:57 <DanC> Of the people who are using it, perhaps you're right, none of them gave de-re/de-dicto any thought. they just ran the code and walked away fat-dumb-and-happy.