Semantic Web Interest Group IRC Chat Logs for 2004-11-15

08:28:04 Topic now Semantic Web / RDF chat n hack. blog http://rdfig.xmlhack.com/ logs http://www.ilrt.bris.ac.uk/discovery/chatlogs/rdfig/

08:28:04 Users on #rdfig: logger Pike_ LotR Talliesin_ grove crschmidt Zenethian eikeon chrisc vdv sh1mmer eaon MarkB ear1grey_ ericP jsled sbp DanC CaptSolo jdougan edd Grantbow hex_ mattmcc dc_rdfig azz dngor workbench Jipp phenny nidd mmealling optika wh4experimental shellac iwaim__ kpreid TrustBot WSBot balbinus6 kota Iter Jibbler GabeW julie sandro karlcow sanctius danbri Cloud_ deltab pjz_ Shammah sandro_ kasei

08:28:04 <NickServ> This nickname is owned by someone else

08:28:04 <NickServ> If this is your nickname, type /msg NickServ IDENTIFY <password>

08:28:04 <NickServ> Password accepted - you are now recognized

08:28:04 <MemoServ> You have no new memos

08:28:04 <ChanServ> [#rdfig] This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

09:23:33 <sbp`> sbp` is now known as sbp

10:08:00 <libby> hello brian!

12:51:40 <Talliesin_> Talliesin_ is now known as Talliesin

13:16:18 <bengee_> bengee_ is now known as bengee

13:27:26 <CloCkWeRX[dc]> CloCkWeRX[dc] is now known as CloCkWeRX

15:18:21 <Jubin> Jubin is now known as jubin|away

16:06:53 <aliman_> aliman_ is now known as aliman

17:09:11 <PhilT> msg danbri You been in touch with Grady Booch yet Dan?

17:09:25 <DanC> :)

17:09:28 <DanC>http://lists.w3.org/Archives/Public/public-cwm-talk/2004OctDec/0015.html

17:09:29 <dc_rdfig> A: http://lists.w3.org/Archives/Public/public-cwm-talk/2004OctDec/0015.html from DanC

17:09:45 <DanC> A:|Notation3 languages, cwm etc: Scheduled topic chat 1900Z

17:09:45 <dc_rdfig> Titled item A.

17:10:30 <DanC> A:see [http://esw.w3.org/topic/ScheduledTopicChat|ScheduledTopicChat], [http://www.w3.org/DesignIssues/Notation3|Notation3]

17:10:30 <dc_rdfig> Added comment A1.

17:11:09 <crschmidt> .t Z

17:11:12 <phenny> 2004-11-15T17:11:10Z

17:12:09 * DanC has invited a few people; hasn't gotten any confirmations

17:32:14 <DanC> A:we'll start with roll call/introductions, build an agenda for a bit, and then more or less focus on one thing at a time. duration around 90min, I think.

17:32:15 <dc_rdfig> Added comment A2.

17:33:13 <DanC> A:attendee [http://www.w3.org/People/Connolly/|Dan Connolly, W3C]

17:33:13 <dc_rdfig> Added comment A3.

17:35:30 * eaon would like to come :/ but can't

17:35:35 <eaon> at least i'm sure you're logging

17:37:40 <DanC> you could let more people know you're interested by writing: A:regrets eaon. will check the log

17:38:11 <DanC> any particular part of the discussion space that you're interested in, eaon ?

17:39:21 <eaon> not yet - still learning :) and i'm not sure it's interesting enough for chump that i'll check the log ;)

17:39:27 <eaon> but i gotta run now - see you

17:46:42 <Talliesin_> Talliesin_ is now known as Talliesin

18:20:13 <DanC> BLURB: cwm ready for 1.0 release? (agendum B)

18:20:14 <dc_rdfig> B: cwm ready for 1.0 release? (agendum B) from DanC

18:21:40 <DanC> B:see [http://www.w3.org/2000/10/swap/doc/plans.html|cwm plans], [http://lists.w3.org/Archives/Public/public-cwm-announce/2004OctDec/0000.html|release candidate 1 announcement], [http://lists.w3.org/Archives/Public/www-archive/2004Nov/0019.html|rc2 build/test log]

18:21:40 <dc_rdfig> Added comment B1.

18:23:53 <crschmidt> oops, looks like i should see if my problem is fixed in recent builds if it's that close to release

18:24:34 <crschmidt> oh, wait, i already did, nevermind.

18:27:55 <DanC> dc_rdfig, view

18:27:57 <dc_rdfig> A: Notation3 languages, cwm etc: Scheduled topic chat 1900Z (http://lists.w3.org/Archives/Public/public-cwm-talk/2004OctDec/0015.html)

18:27:58 <dc_rdfig> B: cwm ready for 1.0 release? (agendum B) (blurb)

18:28:10 * DanC demos dc_rdfig for yosi_

18:31:55 <yosi_> A: yosi_ plans to attend

18:31:57 <dc_rdfig> Added comment A4.

18:33:13 <DanC> A1:see [http://esw.w3.org/topic/ScheduledTopicChat|ScheduledTopicChat], [http://www.w3.org/DesignIssues/Notation3|Notation3], [http://www.w3.org/2000/10/swap/doc/cwm|cwm]

18:33:14 <dc_rdfig> Replaced comment A1.

18:51:37 <DanC> BLURB: Notation3 identifier syntax: what characters are allowd? (agendum C)

18:51:37 <dc_rdfig> C: Notation3 identifier syntax: what characters are allowd? (agendum C) from DanC

18:52:05 <DanC> C:| Notation3 identifier syntax: what characters are allowed? (agendum C)

18:52:05 <dc_rdfig> Titled item C.

18:53:36 <timbl> C:Attendee [http://www.w3.org/People/Berners-Lee/|Tim Berners-Lee]

18:53:37 <dc_rdfig> Added comment C1.

18:54:10 <timbl> A:Attendee [http://www.w3.org/People/Berners-Lee/|Tim Berners-Lee]

18:54:10 <dc_rdfig> Added comment A5.

18:54:15 <timbl> C1:""

18:54:16 <dc_rdfig> Deleted comment C1.

18:54:19 <DanC> C:see bugs reports such as [http://lists.w3.org/Archives/Public/www-archive/2002Feb/0025.html|Notation3: The Great QName Survey]

18:54:19 <dc_rdfig> Added comment C1.

18:54:42 <sbp> A:Attendee [http://inamidst.com/sbp/|Sean B. Palmer].

18:54:43 <dc_rdfig> Added comment A6.

18:54:46 <phenny> sbp: 15:20Z <Morbus> tell sbp to read his email. and that slogger rules.

18:54:51 <phenny> sbp: 16:40Z <xover> tell sbp For sqlwhois->miscoranda s/(left.+?)left/\1right/.

18:55:34 <sbp> thanks phenny

18:55:37 <phenny> Why, don't mention it.

18:59:29 <DanC> BLURB: Notation3 formal grammar (agendum D)

18:59:30 <dc_rdfig> D: Notation3 formal grammar (agendum D) from DanC

18:59:57 <DanC> ah... my clock reads 1900Z

19:00:19 <timbl> D: See [n3.n3|http://www.w3.org/2000/10/swap/grammar/n3.n3]

19:00:19 <dc_rdfig> Added comment D1.

19:00:24 <DanC> who's here for the cwm/n3 chat? if you would, please write: A: Fred Smith attending

19:00:38 <AndyS> A: Attendee [http://www.hpl.hp.com/personal/afs/|Andy Seaborne]

19:00:39 <dc_rdfig> Added comment A7.

19:01:07 <DanC> I'm thinking of a 90 minute duration; anybody like that idea, or prefer something else?

19:01:09 <crschmidt> A: Watching [http://crschmidt.net/|Christopher Schmidt]

19:01:09 <dc_rdfig> Added comment A8.

19:01:17 <dajobe> A:Dave Beckett, here but eating dinner

19:01:17 <dc_rdfig> Added comment A9.

19:01:44 * DanC missed an opportunity to ask folks to note where they're writing from; perhaps another time.

19:02:08 <DanC> so timbl and I have suggested a few items, via the chump...

19:02:10 <DanC> dc_rdfig, view

19:02:10 <dc_rdfig> A: Notation3 languages, cwm etc: Scheduled topic chat 1900Z (http://lists.w3.org/Archives/Public/public-cwm-talk/2004OctDec/0015.html)

19:02:11 <dc_rdfig> B: cwm ready for 1.0 release? (agendum B) (blurb)

19:02:12 <dc_rdfig> C: Notation3 identifier syntax: what characters are allowed? (agendum C) (blurb)

19:02:13 <dc_rdfig> D: Notation3 formal grammar (agendum D) (blurb)

19:02:25 <timbl> BLURB: Streaming RDF a la cwm --pipe -- should @prefix be allowed anywhere in a file?

19:02:26 <dc_rdfig> E: Streaming RDF a la cwm --pipe -- should @prefix be allowed anywhere in a file? from timbl

19:02:57 <gk> A: Graham Klyne, present and almost correct (or something)

19:02:58 <dc_rdfig> Added comment A10.

19:03:24 <timbl> E:| Streaming RDF a la cwm --pipe -- should @prefix be allowed anywhere in a file? (Addendum D)

19:03:25 <dc_rdfig> Titled item E.

19:03:35 <timbl> E:| Streaming RDF a la cwm --pipe -- should @prefix be allowed anywhere in a file? (Agendum D)

19:03:36 <dc_rdfig> Titled item E.

19:04:07 <gk> Am I missing something, Chump A: seems to be attendees, what are the others?

19:04:14 <DanC> I'd like to see comments like "yes, let's discuss those things" or "I'm particularly interested in n3.n3" or "I've got another one: pychinko" or some such.

19:04:21 <sbp> gk: agenda items

19:05:01 <gk> I'd be interested to geta definitive interpretation of :-

19:05:08 <DanC> me too...

19:05:16 <DanC> BLURB: Notation3 :- idiom

19:05:17 <dc_rdfig> F: Notation3 :- idiom from DanC

19:05:25 <DanC> F:| Notation3 :- idiom (agendum F)

19:05:25 <dc_rdfig> Titled item F.

19:05:41 <sbp> there are always syntax points to discuss. your note to me last week that the meeting'd take place (thanks, btw) was over the choice of path characters--general disagreement

19:05:42 <AndyS> I noted "What should the standardization track be if any for N3?" in the call-for-chat

19:05:43 <timbl> If you are new to this chat, refresh http://rdfig.xmlhack.com/ in your browser to get the meeting status.

19:06:02 <DanC> BLURB: Notation3 standardization (item G)

19:06:03 <dc_rdfig> G: Notation3 standardization (item G) from DanC

19:06:04 <AndyS> and also "SPARQL and N3"

19:06:17 <gk> F: [GK] I use it (:-) to create an alias for whatever expression follows; especially the head of a list

19:06:17 <dc_rdfig> Added comment F1.

19:06:29 <timbl> I was thinking identifiers characters... path characters is a choice too.

19:06:38 * DanC wonders if yosi_ is tuned in

19:06:54 <sbp> can tack it onto Agendum C then

19:06:57 <gk> I'm also interested in nailing down identifier characters, especialiiy to '-' or not to '-'

19:07:04 <timbl> Me too.

19:07:24 <dajobe> C:[http://www.ilrt.bris.ac.uk/discovery/2004/01/turtle/#sec-qnames|Turtle issues on qnames]

19:07:24 <dc_rdfig> Added comment C2.

19:07:25 <timbl> I am also getting a definitive list of any syntax issues oustatnding.

19:07:25 <DanC> yes, gk, that's item C

19:08:04 <sbp> character encoding differences was one mentioned in the discussion announcement

19:08:07 * yosi_ tunes in

19:08:19 <sbp> more explicitly: "relationship with NTriples (UTF-8 vs ASCII, etc)"

19:08:35 <AndyS> i18n-ization of QNames would be good

19:08:36 <DanC> ok, noting that the IRC medium allows anybody to comment on anything at any time without too much disruption, let's see... I want yosi's attendion on B

19:09:01 <DanC> so let's do C (name syntax) and timbl, pls get ready to present n3.n3

19:09:15 <sbp> I've been wondering about N-Triples's double encoding for URIs. you can both hex encode utf-8 encoded Unicode, or you can use the \uHHHH|\UHHHHHHHH escapes

19:09:31 <timbl> Should we discuss how long the meeting is now or at the end?

19:09:35 <DanC> DanC has changed the topic to: Notation3/cwm chat http://rdfig.xmlhack.com/ identifier syntax (agendum C)

19:09:37 <sbp> heh

19:09:38 <dajobe> that's at different levels, sbp

19:09:52 <dajobe> N-Triples has just one way to write down ASCII URIs

19:10:06 <timbl> Yes.

19:10:10 <DanC> I suggested 90 minute duration and asked for endorsement and got none, timbl

19:10:19 <dajobe> 90s ok for me

19:10:42 <timbl> I had only blocked out 60..

19:10:44 <DanC> tx

19:10:46 <sbp> sounds like you're suggesting that the unicode characters should be preserved if they appear, but then also that URIs don't allow (unencoded) unicode characters, which is correct at the moment of course

19:10:51 <AndyS> 60 or 90mins OK here : I am "at risk" after 60 mins

19:11:10 <sbp> whereas I'd been under the assumption that if they were allowed by the syntax, you'd have to do something with them (utf-8 and hex encoding) if you found them

19:11:23 <gk> G: I think "stabilization is maybe a more worthy goal than "standardization"; don't want to cramp further experimentation, etc. Also, there might be I18N issues if N3 were to go REC track.

19:11:23 <dc_rdfig> Added comment G1.

19:11:27 <gk> G: I think "stabilization is maybe a more worthy goal than "standardization"; don't want to cramp further experimentation, etc. Also, there might be I18N issues if N3 were to go REC track.

19:11:28 <dc_rdfig> Added comment G2.

19:11:35 <sbp> but I'm getting ahead of myself. B is 1.0RC2

19:12:02 * gk oops, how do I delete a superfluous chump comment

19:12:04 <sbp> which I've just downloaded, and it passes my "echo | python2.4 cwm.py" text, so I'm happy!

19:12:18 <dajobe> gk: do G2:""

19:12:18 <DanC> hmm... I might have a conflict after 60min

19:12:19 <sbp> gk: A1:""

19:12:26 <gk> g2:""

19:12:32 <crschmidt> Captial G

19:12:44 <gk> G2:""

19:12:44 <dc_rdfig> Deleted comment G2.

19:12:50 * gk thanks all

19:13:08 <DanC> C is really a lead-in to D... timbl, can you show us what the formal grammar says about identifier syntax?

19:13:33 <timbl> Ummm.... it is kinda weak on that.

19:13:46 <DanC> hence this discussion :)_

19:14:07 <timbl> It deals with "alphas" and in fact the tokenizer has a list of characters which are *not* identifiers.

19:14:17 <sbp> qname: "(([a-zA-Z_][a-zA-Z0-9_]*)?:)?([a-zA-Z_][a-zA-Z0-9_]*)?" - n3.n3

19:14:20 <dajobe> for Turtle, I went with what cwm implemented, rather than what N3 defined

19:14:26 <sbp> (viz. http://www.w3.org/2000/10/swap/grammar/n3.n3)

19:14:32 <timbl> Foloowing the XML1.1 style of assuming that new character s will be identifier characters in unicode.

19:14:41 * DanC found the same qname rule that sbp found

19:15:07 <gk> The idea of @prefix anywhere rather appeals to me. I use a derivation of N3 for the scripting language in Script -- re-uses a lot of parser software components, etc. The idea of @prefix anywhere fits kida nicely with that, I think.

19:15:20 <timbl> Yes, that is a placeholder. In fact I think Turtle makes reference to the XML spec.

19:15:42 * DanC supposes that=the qname: rule

19:16:11 <timbl> yes. The qname rule should allow unicode characters.

19:16:20 <gk> D:The idea of @prefix anywhere rather appeals to me. I use a derivation of N3 for the scripting language in Script -- re-uses a lot of parser software components, etc. The idea of @prefix anywhere fits kida nicely with that, I think.

19:16:20 <dc_rdfig> Added comment D2.

19:16:21 <dajobe> the research I chumped at C2 for turtle & qnames is what I'd feel looks ok

19:16:25 <sbp> you could just copy the production from XML 1.1 then

19:16:39 <AndyS> Jena does XML 1.1 minus characters that are used for other things, like "_" - not a perfect

19:16:51 * DanC looks at turtle issues http://www.ilrt.bris.ac.uk/discovery/2004/01/turtle/#sec-qnames

19:17:02 <sbp> _ is used for other things? in the name part?

19:17:25 <yosi_> n3.n3 is currently debugged using predictiveParser.py, which has far bigger problems with unicode than what n3.n3 has

19:17:29 <gk> C:I have sometimes found the lack of '-' in N3 qnames to be an inconvenience. It seems at odds with XML.

19:17:29 <dc_rdfig> Added comment C3.

19:17:32 <timbl> - is arguable, but I never saw any reason to exclude _

19:17:53 <DanC> C:whether '-' is allows in N3 or not is definitely worth deciding explicitly in a test case

19:17:54 <dc_rdfig> Added comment C4.

19:17:54 <sbp> IIRC the original spec didn't allow _ to start name productions, but DanC kept using it to make temporary properties: e.g. :_firstName

19:18:07 <DanC> C4:whether '-' is allowed in N3 names or not is definitely worth deciding explicitly in a test case

19:18:07 <dc_rdfig> Replaced comment C4.

19:18:10 <AndyS> _:a is a bNode and a legal qname in XML 1.1!

19:18:28 <AndyS> (if I read the rec right)

19:18:37 <sbp> N3 != XML 1.1

19:18:43 <AndyS> :-)

19:18:54 <DanC> C:other interesting cases: _:a is a bNode and a legal qname in XML 1.1! says AndyS; :_firstName is an idiom I use when I'm lazy.

19:18:54 <dc_rdfig> Added comment C5.

19:19:19 <sbp> N3 says _ is to be treated specially, unless it's bound to @prefix. actually, I'm not fully sure whether @prefix override of that character is intentional--DanC? timbl?

19:19:25 <dajobe> C:also interesting - is ':' a (good/allowed) qname?

19:19:25 <dc_rdfig> Added comment C6.

19:19:27 <timbl> C: I have found the preence of "-" in qnames to be a pain because (1) it means you can't extend the language with math experssions without adding whitespaces;(b) you get this arbitrary question fo whether to use - or _ (or just camelCase) which gives an extra dimension to identifier arbitrariness

19:19:28 <dc_rdfig> Added comment C7.

19:19:34 * DanC finds it hard to care a lot about this issue; wonders whether to aim for a decision in this meeting or a path going forward

19:20:30 <sbp> CWM turns:

19:20:30 * AndyS wants to cover the rest of the agenda and not get into a syntax loop

19:20:31 <sbp> @prefix _: <#> .

19:20:31 <sbp> _:p _:q _:r .

19:20:36 <sbp> into:

19:20:36 <sbp> @prefix : <#> .

19:20:37 <sbp>

19:20:37 <sbp> :p :q :r .

19:21:12 <DanC> I'd like to move on too; the input we got on the name syntax was valuable, yes timbl?

19:21:13 <gk> timbl, (re. '-' in qnames) I appreciate your position. FWIW, the Haskell language has a similar concern with '.' (used as part of module-qualified name, and is also operator in its own right). The distinction is managed lexically -- '.' can be used as part of a name (but not initial char), otherwise is the operator.

19:21:53 <sbp> I'd just like to note that I really enjoy the binary hack to convert - and _ sequences to mere _s

19:21:58 <timbl> Well, I'll use that as a transition to discusiion of CVG for n3

19:22:07 <sbp> (but I don't think it'd be good to deploy it)

19:22:17 * DanC tries to defrobulate CVG; fails

19:22:18 <gk> CVG?

19:22:22 * timbl notes questionaire on syntax http://www.w3.org/2002/09/wbs/1/RDF-N3-Syntax/results

19:22:39 <timbl> sorry CFG

19:22:45 * jsled notes ACL'ed link.

19:22:51 <DanC> C:see also [http://www.w3.org/2002/09/wbs/1/RDF-N3-Syntax/|Syntax of QNames in N3] questionaire

19:22:51 <dc_rdfig> Added comment C8.

19:23:15 <timbl> So there was a plan to have a definition of N3 in RDF.

19:23:18 <DanC> DanC has changed the topic to: otation3/cwm chat http://rdfig.xmlhack.com/ Notation3 formal grammar (agendum D)

19:23:20 <gk> timbl, the link http://www.w3.org/2002/09/wbs/1/RDF-N3-Syntax/results appears to be member-only

19:23:42 <DanC> that would explain why so few results. phpht.

19:23:46 <timbl> The udea being taht no one should moan that it wasn't properly defined, and that we could test the tests against the grammar, and parsers against the tests.

19:23:56 <DanC> logger, chump D

19:23:56 <DanC> D:See [http://www.ilrt.bris.ac.uk/discovery/chatlogs/rdfig/2004-11-15#T19-23-56|discussion]

19:23:57 <dc_rdfig> Added comment D3.

19:24:04 * timbl member only? shouldn't be! "open to the public"

19:24:29 <sbp> definitely 401

19:24:30 <timbl> ACTION TimBL: figure out why people can't access it later

19:24:47 <sbp> which "group" does it pertain to?

19:24:57 <timbl> So, it turns out that doin the context-free grammar in RDF was quite fun.

19:24:58 <yosi_> B:Does anyone have any bugs / comments / concerns with the cwm 1.0 release candidate?

19:24:59 <dc_rdfig> Added comment B2.

19:25:05 <gk> I didn't know it was there... where was it announced?

19:25:16 * sbp neither

19:25:26 <timbl> e cfg:mostBeOneSeqeunec ( ( a b ) ( c d ) ).

19:25:36 * AndyS was asked for password for WBS - accessing from non-HP network

19:25:39 <timbl> I think it was mentioned here

19:25:41 * DanC sent a sysreq re http://www.w3.org/2002/09/wbs/1/RDF-N3-Syntax/ , cc timbl

19:25:47 <timbl> tx

19:26:11 <sbp> yosi_: can you please send a quick note about CWM 1.0RC2 to public-cwm-announce, please? even a single line to say it's out there would be good

19:26:21 <DanC> gk, the cwm1.0rc1 announcement is linked from B; it went to public-cwm-announce

19:26:24 <timbl> that one property allows you to say "This producion muse expand as one of the following sequenxces of productions".

19:26:31 <sbp> and did you skip 0.9.*?

19:26:35 <yosi_> yes

19:26:38 <DanC> sbp, that's agendum B; can we get to that after the formal grammar?

19:26:51 <sbp> you want me to get back to B after C? sure :-)

19:27:02 <timbl> In fact I used fomr shortcuts like [ bnf:zeroOrMore declaration ]

19:27:13 * DanC re-reads... is happy for yosi to announce rc2 meanwhile

19:28:21 * DanC wonders if timbl is typing or waiting

19:28:28 <timbl> N3 happens to be a very simple language (by design) , specifically though one which cna be parsed by looking ahead one token only to decide which branch of a production to expand a production as.

19:28:30 <sbp> yep, the RDF BNF seems a major success. I'd really like to hook it up to a proper recursive-descent toolkit, i.e. create an N3 parser in Python direct from the N3 N3 grammar

19:28:37 <adrianw> Can someone point me to a copy of the agenda please

19:28:48 <sbp> (actually, I suppose the RDF/XML should be canonical)

19:28:51 <DanC> welcome, adrianw . agenda: http://rdfig.xmlhack.com/

19:28:55 <timbl> Well, I did make a python parser driven from the grammar.

19:29:05 <sbp> you did? great! is it on the Web?

19:29:34 <gk> sbp, my parser (both of them) was constructed something like that -- i.e. hand-coded recursive descent from syntax

19:29:41 <timbl> adrianw, load (and frequently refresh) rdfig.xmlhack.com (see topic)

19:30:04 <sbp> well all have been hand coded from the syntaxes, even the yapps and spark stuff...

19:30:15 <timbl> sbp, tes: http://www.w3.org/2000/10/swap/grammar/predictiveParser.py

19:30:18 <timbl> yes

19:30:21 <sbp> but the RDF BNF is machine readable, so no need to even transcribe

19:30:28 * sbp looks

19:30:38 <DanC> D:see [http://www.w3.org/2000/10/swap/grammar/Makefile|grammar/Makefile] for many details

19:30:38 <dc_rdfig> Added comment D4.

19:30:49 <timbl> Actually it reads in teh BNF after it has been decorated by some ruels.

19:31:05 <timbl> The rules track what character each production can possibly begin with.

19:31:12 <gk> sbp, gotcha, misunderstood your comment.

19:31:33 <DanC> hmm... I thought check-grammar.py got renamed to predictiveParser.py ; Makefile still shows check-grammar.py

19:32:21 <DanC> yosi, can you help us find the part of the test log (http://lists.w3.org/Archives/Public/www-archive/2004Nov/0019.html) that regards the formal grammar?

19:32:24 <timbl> Yes, check the makefile. The file IIRC is n3-selectors.n3 which contains a mapping for each production from next character to which branch to take. predictiveParser.py loads taht and then parses the file of your chocie based on that.

19:32:40 <timbl> It only checks the sysntax.

19:32:50 <timbl> It doesn't generate triples.

19:33:15 <timbl> Also, the n3.n3 file doesn't contain the tokenizing, which contains the @keywords magic.

19:33:32 <DanC> D:I hope to obsolete notation3.py in favor of this work soonish; 1.1? but for now, the formal stuff only checks syntax; doesn't call the API yet.

19:33:32 <dc_rdfig> Added comment D5.

19:33:44 <timbl> But the result is that we can be pretty sure in the long run that the tests match the definition.

19:34:06 <timbl> D: No work done to test speed of parser.

19:34:07 <dc_rdfig> Added comment D6.

19:34:18 <timbl> D: Also generates a yacc file

19:34:19 <dc_rdfig> Added comment D7.

19:34:33 <DanC> yosi, the test log doesn't show grammar tests; didn't you tell me the n3.n3 grammar is tested during pre-release?

19:34:50 <gk> I wonder how good the error recovery/diagnostics are for automatically generated parsers...

19:34:51 <timbl> So this is partly why now more focus on typing up loose ends in the grammar.

19:35:06 <DanC> the one diagnostic I saw was very good, gk.

19:35:08 <yosi_> look at test 261/283

19:35:12 <AndyS> Would be good grammar to turn into "member submission" style fixed page

19:35:18 * timbl wishes he could find DanC's experssion of surprsie at the nice error message ;-)

19:35:19 <yosi_> through 283/283

19:35:28 <sbp> "20041105: predictiveParser.py does all the positive parser tests yosi has come up with" - http://www.w3.org/2000/10/swap/doc/plans

19:35:29 <DanC> "261/283) n3parser.tests_n3_10001.out An N3 file with no final newline character"

19:35:30 <DanC> ah; thanks.

19:36:04 <timbl> It doesn't do error recovery.

19:36:33 <AndyS> gk - pretty good - and usually can add special checking for even better message (Jena uses Antlr - an LL parser - but nothing special on parse errors - it gets line/col right)

19:36:48 <timbl> n3.n3 got cleaned up, a bit more constrained. Like the declarations at the top.

19:36:52 <gk> Actualloy, I didn't really mean recovery so much as making a r4asonable attempt to get back on track so as not to produce zillions of secondart syntax errors.

19:37:03 <DanC> my WG-chair instincts have me starting to recruit somebody to reproduce timbl's work on the grammar, but we don't necessarily have a next meeting. :)

19:37:42 <sbp> wouldn't mind making it a monthly or sporadic thing

19:37:50 <DanC> timbl, you seem to have explained n3.n3 well enough that a few people understood. shall we move along?

19:37:51 <timbl> Objective: To get N3 and turtle ligned up very closelyt

19:37:59 <DanC> ah

19:38:13 <timbl> I think lots of people understand it, yes.

19:38:31 <timbl> I think that one thing it opens up is demoantrating that turtle is a sublanguage of N3.

19:38:37 <DanC> turtle is now official DAWG business... trying to remember who has the action... nobody on record, I don't think, though DaveB will have to be involved, of course.

19:38:42 * yosi_ looks up lign in a dictionary

19:38:43 <timbl> (or not, and fixng it)

19:38:45 <sbp> oh, did you fix the optional trailing period in formulae bug?

19:39:00 <AndyS> Turtle is official DAWG business?

19:39:10 * DanC helps yosi defrobulate: s/ligned/lined/

19:39:12 <sbp> the thing that was keeping N3 from being LL(1)

19:39:29 <DanC> yes, the WG resovled that the test materials shall use turtle, recall?

19:39:59 <timbl> Does that mean that the WG is planning to standardize turtle?

19:40:37 <DanC> "that tests shall have input in RDF/XML and turtle ... RESOLVED" -- http://lists.w3.org/Archives/Public/public-rdf-dawg/2004OctDec/0210.html

19:40:49 <timbl> sbp, the current n3.n3 has an optional trailing period. is predictive-parsable which I think is a subset of ll1

19:40:51 <DanC> inasmuch as RDF Core standardized n-triples, yes, I believe so.

19:41:00 <sbp> thanks

19:41:13 <AndyS> I do not read "Use turtle" as "standardize" (a matter of time)

19:41:24 <AndyS> But that's off track

19:41:46 <timbl> So if tuutle and N3 are to stay alligned, how will that be managed? Should those intersted in N3 join DAWG to discuss thinsg liek identifier syntax?

19:41:47 <DanC> yeah, I dunno whether we'll do a whole test WD and all that

19:42:24 <AndyS> I have concerns here.

19:42:33 <DanC> I think the overlap between those here and DAWG is already plenty, timbl

19:43:31 <timbl> dajobe, You made a turlle.n3 a la n3.n3, didn't you, and sent it to a list?

19:43:37 * DanC realizes we're blurring into Notation3 standardization, and Notation3/DAWG... wonders if we should go back and do B now, checking the clock

19:43:42 <dajobe> yeah, I sent it to www-archive

19:43:49 <dajobe> seems we're on G, not D

19:43:58 <timbl> I had an writted todo to try it against your tests using predictiveParser.py

19:44:17 <dajobe> it's at http://lists.w3.org/Archives/Public/www-archive/2004Nov/0004.html

19:44:19 <sbp> it would be nice if Turtle were specified in such a way as to be extensible to being a subset of Notation3; not sure what that entails though. for example, if it had an RDF BNF, supported being modularised... it feels like Turtle should be profiled from N3, not the other way around

19:44:38 <timbl> Some of the comments on "G" maybe need to be tacked onto it in the chumpog

19:44:46 <DanC> D:interesting... [http://lists.w3.org/Archives/Public/www-archive/2004Nov/0004.html|Approximate turtle in BNF n3]

19:44:46 <dc_rdfig> Added comment D8.

19:44:56 <DanC> logger, chump G

19:44:56 <DanC> G:See [http://www.ilrt.bris.ac.uk/discovery/chatlogs/rdfig/2004-11-15#T19-44-56|discussion]

19:44:57 <dc_rdfig> Added comment G2.

19:45:12 <dajobe> D:not annotated there are the things I took outof n3.n3 to make turtle.n3

19:45:13 <dc_rdfig> Added comment D9.

19:45:20 <sbp> dajobe, answering my points before I even ask them as usual :-)

19:45:38 <DanC> Andy, I'm curious about your concern. care to elaborate?

19:46:04 <timbl> I think predictive parsable languages are simple to show equivalence.... just gut feeling.

19:46:08 * DanC nominates sbp to organize a future scheduled topic chat, and/or chair the tail end of this one

19:46:20 <AndyS> Not here - not-technical

19:46:26 <DanC> very well

19:46:27 <sbp> having to go already, DanC?

19:46:34 <DanC> TAG meets at xx:00

19:46:43 <sbp> ah, so timbl off too?

19:46:51 <DanC> we may be able to divide our attention

19:46:52 <timbl> Yes, unless I desert it

19:47:16 <DanC> DanC has changed the topic to: Notation3/cwm chat http://rdfig.xmlhack.com/ cwm ready for 1.0 release? (agendum B)

19:47:31 <DanC> first, a round of applause for yosi's release engineering work, please

19:47:42 <sbp> yep, well done yosi

19:47:47 <sbp> lots of forward momentum lately

19:47:47 * yosi_ blushes

19:47:54 <AndyS> Well done, Yosi

19:48:00 <gk> cheers

19:48:00 * timbl both ... hmm That's true. Multiple windows. But I'll have to stop listening to the Bach violin sonatas. ;)

19:48:16 <timbl> yes, Yosi has done great things.

19:48:19 <JosD> good good

19:48:20 <sbp> is Agendum B resolved--going to 1.0? nobody had complaints when yosi asked

19:48:21 <DanC> sbp, you said rc1 passed your smoke-test, yes?

19:48:25 <sbp> aye

19:48:43 <sbp> RC2 in fact

19:48:43 <timbl> (Just wading in and figuring out what was going on was medal-worthy!)

19:48:53 * DanC doesn't think we got everybody's attention when yosi asked

19:49:00 <timbl> smoke?

19:49:25 <timbl> Install it on sbp's machine and se if it goes up in smoke?

19:49:37 <DanC> I think the smoke test was: echo | python cwm.py

19:49:45 <crschmidt> The only problem I ran into at all has been fixed in the most recent release.

19:49:48 <sbp> right. if more than 50% of the box remains, I declare it a success

19:50:02 <sbp> but yep. I noted it to public-cwm-bugs too

19:50:05 <DanC> good to know, crschmidt

19:50:30 <DanC> so, anybody think we should get more feedback before calling it 1.0?

19:50:32 <crschmidt> (Granted, I've only used cwm about 5 times total, and usually just for ntriples/n3->rdf/xml translation or the other way around)

19:50:55 <timbl> (going)

19:51:00 <timbl> (going)

19:51:04 <sbp> I'd like to put it to some major project use before I'd be really happy with it. thinking about this like a CR status

19:51:19 <sbp> but I'd be happy if you said "no, it's just a number--go ahead"

19:51:29 <AndyS> "Release early, release often"

19:51:33 <sbp> yeah

19:51:37 <DanC> yes, this will still be "a release with a zero on the end", sbp

19:51:43 * sbp nods

19:51:50 <yosi_> the `smoke test' reminds me of http://simson.net/photos/hacks/cubefire.html

19:51:50 <timbl> CR and 1.0 both mean "go ahead and use it" in a way.

19:51:53 <gk> Yeah, it's been a few years, must be about time foir V1.0 ??

19:52:30 <DanC> B: sentiment from several is: yes, seems ready for 1.0

19:52:30 <dc_rdfig> Added comment B3.

19:52:34 * timbl lol cubefire ... ah NeXTStep! the godo old days :)

19:52:52 <sbp> heh, heh

19:52:57 <DanC> yosi, did you get a log including the reification tests?

19:53:29 <sbp> for Agendum C, can we say that we'll await the outcome of the questionnaire once timbl fixes the ACLs on it? the item expires at the end of this year, IIRC, so one and a half months. is that too long to wait?

19:53:32 <DanC> adrianw, were you able to make heads or tails of any of this? we may have been assuming a lot of shared context.

19:53:33 <yosi_>http://lists.w3.org/Archives/Public/www-archive/2004Nov/0020.html

19:53:38 <dc_rdfig> H: http://lists.w3.org/Archives/Public/www-archive/2004Nov/0020.html from yosi_

19:54:02 <DanC> H:|a rebuild of cwm 1.0.0rc2 log

19:54:02 <dc_rdfig> Titled item H.

19:54:09 <dajobe> H:that contains some un-checked-in CVS changes

19:54:09 <dc_rdfig> Added comment H1.

19:54:22 <adrianw> DanC -- I'm learning fast (I think(:-))

19:54:29 <DanC> ok, good to know.

19:54:37 <DanC> yosi_, help me find the reification tests?

19:55:03 <sbp> E:|Streaming RDF a la cwm --pipe -- should @prefix be allowed anywhere in a file? (Agendum E)

19:55:03 <DanC> holy cow, dajobe, how did you find those so fast?

19:55:03 <dc_rdfig> Titled item E.

19:55:06 <sbp> (mistitled)

19:55:09 <DanC> oh... the M's at the top?

19:55:13 <dajobe> DanC: yes

19:55:17 <yosi_> dajobe, it tried to commit at the end --- and failed

19:55:28 <yosi_> due to DanC modifying a file

19:55:46 <yosi_> I had to commit anyway right afterward

19:55:56 <DanC> H:750 point bounty for anybody who can reproduce yosi's results independently.

19:55:56 <dc_rdfig> Added comment H2.

19:56:16 <sbp> E:<gk> The idea of @prefix anywhere rather appeals to me. I use a derivation of N3 for the scripting language in Script -- re-uses a lot of parser software components, etc. The idea of @prefix anywhere fits kida nicely with that, I think.

19:56:16 <dc_rdfig> Added comment E1.

19:56:22 <sbp> D2:""

19:56:23 <dc_rdfig> Deleted comment D2.

19:56:41 <yosi_> DanC, 111/297 starts reification tests

19:56:47 <timbl> Lots of things left to discuss ... N3 test file manifests

19:56:53 <DanC> "111/126 reify/detailed.tests#t1004 reify twice"

19:56:55 <DanC> ah. thx.

19:57:11 <timbl> Things get very interconnected.

19:57:15 <DanC> what about N3 test file manifests?

19:57:22 <sbp> C:Can we resolve this solely through the questionnaire? Is everybody just assuming that now and I'm the only one left wondering?

19:57:22 <dc_rdfig> Added comment C9.

19:57:40 <timbl> Yosi generaetd a set of N3 positive parser tests from a asubset ofth e original cwm tests.

19:57:57 <DanC> sbp, I don't know

19:58:19 <sbp> the only issue I can see is Turtle/N3 alignment. if N3 diverges from XML 1.1 and Turtle uses that...

19:58:20 <timbl> people have experssed fairly strong feelings on either direction.

19:58:56 <DanC> G:a Notation3 test suite with multiple independed implementations is an interesting goal

19:58:57 <dc_rdfig> Added comment G3.

19:59:07 <DanC> G3:a Notation3 test suite with multiple independent implementations is an interesting goal

19:59:08 <dc_rdfig> Replaced comment G3.

19:59:08 <timbl> I don't see that N3 needs to be bound by XML qnames to that extent, because there is always a way of using <> for problem cases.

19:59:09 <dajobe> sbp: turtle doesn't use xml1.1 yet.

19:59:23 <sbp> ah, okay

19:59:30 <timbl> yet

19:59:32 <timbl> ?

19:59:49 <gk> Re. comment G3, that would be nice. I'd like to bring my Haskell implemenation into line.

19:59:50 <dajobe> I'm not pushing it either way, wait till you think about it for n3.

20:00:11 <timbl> So is the feeling here that the XML 1.1 set should be used as a basis, )and then "." and maybe "-" removed)

20:00:35 <timbl> gk: cool.

20:00:43 <DanC> G:we have cwm, gk's hasell stuff [pointer?], a Jena parser, dajobe's turtle stuff

20:00:43 <dc_rdfig> Added comment G4.

20:00:58 * sbp goes to see if jcowan is around... rats, he isn't

20:01:03 <dajobe> xml namespaces 1.1, since you remove the ':' too

20:01:06 <AndyS> I have some users who are not English speakers - using <> for any URIs that contain accented chars is not nice esp. as RDF/XML can do it.

20:01:23 <gk> timbl, I could certainly live with that, but I wonder if this creates some present inconvenience for the sake of possible future benefits?

20:01:37 <DanC> my attention is divided from now on... any offers to organize a next meeting?

20:02:05 <DanC> ... or offers to summarize this one in email? (sbp?)

20:02:09 <AndyS> There has been a N3-rules version of Jena - not in the main codebase.

20:02:21 <sbp> certainly. I'll do both if you like

20:02:25 <timbl> hmmm http://www.w3.org/2000/10/swap/grammar/predictiveParser.py uses notQNameChars = singleCharacterSelectors + "@" # Assume anything else valid qname :-/

20:02:28 <timbl> notNameChars = notQNameChars + ":" # Assume anything else valid name :-/

20:02:43 <sbp> though organization by consensus for the next meeting, on public-cwm-talk, may be a better route for organising the next meeting

20:02:45 <timbl> sbp, thank you!

20:02:58 <gk> My Haskell stuff, frontpage is here: http://www.ninebynine.org/RDFNotes/Swish/Intro.html, parser is here: http://www.ninebynine.org/Software/HaskellRDF/RDF/Graph/N3Parser.hs

20:03:05 <timbl> (or a questionaire?... maybe when we get them accessible)

20:03:10 <DanC> A:next meeting: sbp to organize

20:03:10 <dc_rdfig> Added comment A11.

20:03:27 <sbp> 19Z on a Monday seems a good time for people; the question seems mainly how long we should wait

20:03:37 <sbp> er, except for TAG. perhaps you should reschedule the TAG meeting :-)

20:03:47 <dajobe> get the TAG to standardise n3, solve 2 problems.

20:03:50 <dajobe> um or gain more

20:03:56 <sbp> heh, heh

20:04:05 <gk> G: My Haskell stuff, frontpage is here: http://www.ninebynine.org/RDFNotes/Swish/Intro.html, parser is here: http://www.ninebynine.org/Software/HaskellRDF/RDF/Graph/N3Parser.hs

20:04:05 <dc_rdfig> Added comment G5.

20:04:16 <DanC> G4:we have cwm, [http://www.ninebynine.org/RDFNotes/Swish/Intro.html|swish] incl [http://www.ninebynine.org/Software/HaskellRDF/RDF/Graph/N3Parser.hs|N3Parser.hs] , a Jena parser, dajobe's turtle stuff

20:04:17 <dc_rdfig> Replaced comment G4.

20:06:11 <DanC> 19Z is late for EU, I gather

20:06:25 <AndyS> G: [http://eulersharp.sourceforge.net/|Euler]

20:06:26 <dc_rdfig> Added comment G6.

20:06:33 <timbl> I think we should aim fro earlier

20:06:34 <sbp> Jos, Andy, and I made it

20:06:53 <timbl> for Europe .. unless we have Asia/Pacific people interested?

20:06:55 <JosD> :)good good thx

20:06:57 <sbp> and Graham and dajobe

20:07:37 <AndyS> G: [http://www.mindswap.org/~katz/pychinko/|Pychinko]

20:07:38 <dc_rdfig> Added comment G7.

20:07:51 <timbl> If this time is fine then OK. Maybe 30 mins ealier to avoid TAG conflict.

20:08:14 <gk> Yup, 19:00 is OK unless I have other engagements (about 1/month on Mondays)

20:09:11 <sbp> G:And other older ones at [http://www.w3.org/DesignIssues/Notation3|Notation3 DesignIssue] (h2: BNF Grammars and parsers)

20:09:11 <dc_rdfig> Added comment G8.

20:10:17 <JosD> G: [http://cvs.sourceforge.net/viewcvs.py/eulermoz/eulermoz/js/parser/n3/|some other ongoing work for Mozdev]

20:10:17 <dc_rdfig> Added comment G9.

20:10:43 <sbp> I spy a JavaScript N3 parser! wow

20:10:45 <jsled> wow.

20:11:52 <timbl> Did I know about that one?

20:12:20 <timbl> I know Dan Jackson had a RDF/XML parser in js.

20:12:27 <sbp> and Jim Ley

20:12:35 <JosD> there's more engine stuff at http://cvs.sourceforge.net/viewcvs.py/eulermoz/eulermoz/js/parser/n3/

20:13:05 * DanC scrolls back to figure out when JosD arrived

20:13:15 <JosD> oops.. I meant http://cvs.sourceforge.net/viewcvs.py/eulermoz/eulermoz/rdfinf/

20:13:44 <JosD> arrived later (traffic jams..) around 7:10 Z

20:14:09 <sbp> what's the goal of EulerMoz, Jos? to try to replace some of the RDF handling mechanisms in Mozilla, or for other RDF projects that people might like to impelement therein?

20:14:41 <gk> G: [http://www.wiwiss.fu-berlin.de/suhl/bizer/rdfapi/|Chris Bizer's RAP stuff in PHP] apparently includes N3

20:14:41 <dc_rdfig> Added comment G10.

20:14:57 <sbp> root documentation seems to be a prolog simulator, so I guess it might be an even wider domain

20:14:59 <DanC> ooh... agenda+ EulerMoz, pls

20:15:29 <sbp> G:RAP just reuses some of my oldest N3 code, IIRC. Shame they didn't use afon!

20:15:30 <dc_rdfig> Added comment G11.

20:15:56 <timbl> Looks as though pychinko uses afon, sbp?

20:16:01 <JosD> sbp: move glue logic (done in glue code, e.g. mvc) to nice inferential stuff :)

20:16:09 <sbp> timbl: yep

20:16:09 <timbl> (Anyone from the pychinko world here?)

20:16:54 <DanC> I thought RAP was PHP, not python

20:16:56 <sbp> they've been quite happy with it, though I think they're planning to write a more up-to-date module at some point. jordan's been working on it--don't think he's around

20:17:03 <sbp> DanC: yeah, they ported it

20:17:04 <JosD> .. still premature http://sourceforge.net/projects/eulermoz and http://eulermoz.mozdev.org/

20:18:02 <sbp> thanks Jos

20:18:39 <gk> Was/is there any emerging consensus re. standardization or not?

20:19:26 <timbl> There wasn't really

20:19:39 * DanC finishes TAG meeting, a quickie for onc

20:19:41 <DanC> once

20:19:48 <sbp> G:No emergent consensus.

20:19:48 <dc_rdfig> Added comment G12.

20:20:04 <DanC> I like the idea of lots of implementations converting on a Notation3 test suite.

20:20:11 <DanC> converging even

20:20:20 <sbp> agreed

20:20:28 <JosD> fully agreed

20:20:36 <gk> I sense several people agree that commonality would be good, is anyone asking for standardization?

20:20:48 <gk> I, too, think a test suite would be good.

20:20:50 <timbl> And the difference?

20:21:03 <timbl> And the difference? between commonality and standardization

20:21:04 <timbl> ?

20:21:06 <DanC> the difference is charter writing, mandate, endorsement, press interaction

20:21:07 <sbp> test suites work in the real world

20:21:26 <timbl> both need test suites and a document

20:21:31 <timbl> and some place to meet.

20:21:32 <AndyS> D:[http://cvs.sourceforge.net/viewcvs.py/jena/jena2/src/com/hp/hpl/jena/n3/n3.g?rev=1.14&view=log|N3 grammar] for [http://www.antlr.org|ANTLR] (Java, C#, C++)

20:21:33 <dc_rdfig> Added comment D9.

20:22:12 <gk> BTW, my test cases are in http://www.ninebynine.org/Software/HaskellRDF/RDF/Graph/N3ParserTest.hs. Maybe not very accessible, right now, but I can haul stuff out and contribute to a common framework if that helps.

20:22:14 <sbp> depends how rigorous the standardisation is to be. I was thinking along the lines of REC-track, which would stall the development quite considerably

20:22:29 <timbl> Yosi, do you have a URI for your list of positive parser tests?

20:22:40 <sbp> if it were to just be a note, or not even that, some central specification would be fine. but then we'd still want to develop a test suite

20:22:47 * AndyS can also drag up his labguage tests

20:23:01 <yosi_> www.w3.org/2000/10/swap/test/n3parser.tests

20:23:41 <DanC> an IG Note seems worthwhile. i.e. the charter-writing and such is already done for the SemWeb IG.

20:23:50 <AndyS> As well as the language, there is the library of builtins.

20:23:55 * JosD wonders where Andy's "Jena with N3 rules" lives??

20:24:00 <AndyS> That would be useful as well.

20:24:20 <sbp> can IGs produce notes?

20:24:25 <DanC> yes

20:24:28 <timbl> Yes, and some liaison with other groups necessary fro builtins.

20:24:46 <yosi_> timbl, it seems that document is squatting on http://www.w3.org/2004/11/n3test , considering that I have no write access there

20:24:49 * DanC wonders about CVS write access to the N3 test suite

20:25:44 <AndyS> Can W3C do a member submission?

20:25:46 <sbp> oh, builtins: that reminds me of a major outstanding RFE I have to request that builtins be more easily extensible by the public... at the moment it's all hardcoded; it'd be nice if CWM browsed a path looking for extension modules. I'll public-cwm-bugs it though

20:25:48 * timbl wonders about a distributed test suite

20:26:10 <DanC> no, but there's a team submission process

20:26:12 <gk> BLURB: Test suites

20:26:12 <dc_rdfig> I: Test suites from gk

20:26:20 <gk> I: [www.w3.org/2000/10/swap/test/n3parser.tests]

20:26:20 <dc_rdfig> Added comment I1.

20:26:40 <sbp> W3C member submitting to itself: it's like rdfs:Class rdf:type rdfs:Class .

20:27:01 <timbl> sbp: builtin path. You can do an "import myBuiltins" in yoru code, and your builtin will register itself with llyn.

20:27:12 <timbl> do you need something more complicated?

20:27:14 <sbp> IIRC, there were two companies that technically we owned by one another at one point

20:27:22 <DanC> where in "yoru code", timbl? doesn't it have to go in cwm.py?

20:27:32 <sbp> timbl: oh, so it's not hard-coded as imports in llyn now?

20:27:45 * sbp peruses the source

20:27:49 <timbl> The default imports are in llyn IIRC.

20:27:55 * timbl looks

20:28:11 <gk> I:[http://www.ninebynine.org/Software/HaskellRDF/RDF/Graph/N3ParserTest.hs|Swish test cases]

20:28:11 <dc_rdfig> Added comment I2.

20:28:42 * gk I thought some more test cases were mentioned too, oh well

20:28:51 <AndyS> I: [http://cvs.sourceforge.net/viewcvs.py/jena/jena2/testing/N3/|Some of Jena's tests for N3] as an RDF language

20:28:51 <dc_rdfig> Added comment I3.

20:29:26 <timbl> spb, yes, myBultins.register(myStore) needs to be called, could be called on module load

20:29:39 <gk> I1: [http://www.w3.org/2000/10/swap/test/n3parser.tests|SWAP tests]

20:29:39 <dc_rdfig> Replaced comment I1.

20:30:22 <sbp> hmm, but that's not possible without modifying the existing CWM code right?

20:30:38 <sbp> that's my aim: to have builtins modules be drop-in-able

20:31:01 <JosD> I: [http://eulersharp.sourceforge.net/2004/01swap/docs/java/index.html|Eulersharp test results]

20:31:01 <dc_rdfig> Added comment I4.

20:31:29 <timbl> Oh, you mean into cwm as a program, not using an API? Sorry. Yes, I suppose one could do that. My current thinking for that sort of thing is to have one environment variable which points to a config file.

20:31:31 <sbp> e.g. when I write a new python module in general, I don't want to have to add a line in the Python source code and rebuild it from source

20:31:36 <sbp> yep

20:31:53 <sbp> env-var would be fine

20:32:09 <timbl> export CWM_CONFIG=~/cwmrc.n3

20:32:15 <timbl> OWTTE

20:32:39 <DanC> if you do an envar, pls do a command-line arg too

20:32:50 <DanC> cwm --config ~/cwmrc.n3

20:33:04 <timbl> And then ... what </home/sbp/lib/cwm/complex.py> a cwm:UserBuiltIn.

20:33:09 <yosi_> at which point timbl, cwmrc.n3 can't use any of our clever extensions, now can it

20:33:35 <sbp> and perhaps </home/sbp/lib/cwm/builtins> a cwm:UserBuiltinsDir .

20:33:44 <sbp> (use all *.py in that folder)

20:33:47 <gk> Do we see special properties, ala CWM, as *the* preferred way of providing built-in style functionality?

20:34:20 <timbl> I have gotten pretty wedded to it.

20:34:21 <DanC> sbp, about when the next meeting happens... which way are you leaning?

20:34:31 <DanC> my schedule is kinda crazy thru 4Dec

20:34:40 <sbp> for 16:30Z in either a couple of weeks or a month

20:34:50 <JosD> I:[http://eulersharp.sourceforge.net/2004/01swap/docs/net/index.html|EulerSharp (C# implementation) test results]

20:34:50 <dc_rdfig> Added comment I5.

20:34:52 <sbp> a month'd be alright

20:34:53 * dajobe will be at ACmtg, W3C@10 fwiw

20:34:57 <timbl> There are optimziations you can do with respect to when you evaluate knowledge to search for and builtins.

20:35:17 <sbp> that'd be 2004-12-13T16:30Z

20:35:23 <d2m> d2m is now known as d2m_

20:35:35 * DanC takes a phone call...

20:36:09 <gk> I've not anything against special properties, but I've experimented with other approaches, mainly using special functions to modify variable bindings as part of a rule definition.

20:36:14 <timbl> 2004-12-13T16:30Z looks OK to me

20:36:46 <gk> I guess at some stage, when rules are standardized, this is something that will become fixed for the community.

20:37:12 * timbl tries to figure out what gk means by 'special functions to modify variable bindings as part of a rule definition."

20:37:35 <sbp> separating tests/builtins from the antecedent is still something I'm wondering about. I asked once what happens when you want to query for something that is also a builtin triple, and the reply was that if the triple is in the store, CWM'll treat the builtin triple in the query as just a regular triple

20:37:41 <timbl> I worry that SPARQL will set a tone of having builtins be special syntax.

20:38:19 <sbp> so now I'm wondering what happens if you've a builtin triple in the store, yet you want the query builtin to act as a builtin. or, even more absurdly, if you wanted some builtins to remain special, and some to be treated as just other triples to query

20:38:23 <AndyS> timbl - example?

20:39:01 <gk> Rather than use an additional property in the antecedent of a rule, the "variable modifier" is applied after the antecedent has been matched, and filters and/or modifies the variable bindings from the antecedent match before substituting into the consequent. It works a bit for backward chaining, too, but not so well.

20:39:02 <timbl> sbp, cwm will search for the tripples in the store as well as trying to claculate them.

20:39:07 <sbp> all of which could be solved if the builtins/tests were separate, as they are in *QL languages--*but*, I really don't like that separation. the current mechanism in N3 is pragmatic, and it's most befitting the RDF model

20:39:20 <sbp> it'll do both? ah... hmm

20:39:45 <sbp> (OTOH, there's N3QL of course)

20:40:19 <sbp> gk: sounds like we're talking about roughly the same thing?

20:40:32 <timbl> Andry, example: SELECT ?x WHERE ?x color ?y AND ?y > ?10" . Why is > not math:greaterThan (syntax apart)

20:41:27 <sbp> of course, I gather that that's how it's implemented now anyway. you just remove all the builtins, do the query as normal, then the builtins after... though I suspect now that CWM does the builtins as it goes along, for efficiency

20:41:35 <timbl> Andy, it may be that an engine can infer ?y > 10 from ?y > z and z > 10. but remobving it from the main question makes it separate, non standard knowledge.

20:42:06 <gk> sbp, possibly the same. I've also experimented with trying to use class-restriction based reasoning (sort-of like OWL), but after some initial success I'm less impressed. So far, performance is lousy.

20:42:26 <sbp> sounds interesting. results of that on the Web?

20:43:30 <timbl> CWM divides builtins inbto light and heavy. Light ones (like adding two integers) are done before any search. Heavy ones (like web access or remote query) are done after local search.

20:43:44 <AndyS> tim - its just surface syntax - if I had my way "?x > ?y" would be defined as "?x math:greaterThan ?y"

20:43:54 <sbp> how are the light ones done before a search if there are no bindings yet?

20:44:08 * DanC looks up 1630Z... find it's 10:30am Chicago time. http://www.timeanddate.com/worldclock/fixedtime.html?month=12&day=13&year=2004&hour=16&min=30&sec=0&p1=0 seems to work for me

20:44:45 <gk> sbp, it's all in the Swish software. It's under the general area of datatype-aware inferencing. I've also done a write-up of some of my explorations: http://www.ninebynine.org/RDFNotes/RDF-Datatype-inference.html

20:44:49 <timbl> The light ones are done iff and whenn :-) there are enough bindings already made.

20:44:58 <sbp> ahh, okay

20:45:02 <sbp> gk: thanks

20:45:20 <sbp> .gc whenn

20:45:29 <AndyS> Also of course the link to F&O + no rules on execution order (nec. as we have people do rewrite to SQL - no order can be assumed)

20:45:39 <sbp> (<phenny> whenn: 7,630)

20:45:54 <AndyS> I thouhgt you would mention SOURCE and log:semantics

20:46:08 <sbp> (<phenny> whenn "when and only when": 5)

20:47:22 <gk> BTW, one thing I did notice is that backward chaining works much better when the special properties are processed along with antecedent matching against the database... which kind-of supports what timbl is describing, I think.

20:48:12 * JosD agrees with gk

20:48:19 <AndyS> Are there examples of "?x math:greaterThan ?y" => SQL engines? Other predicates? given the light/heavy rules

20:48:21 <timbl> I'd love for SPARQL to have log:semantics of course, but not in 1.0 probably.

20:49:00 <AndyS> SOURCE is different and can't be done in cwm :-( as far as I can see.

20:49:05 <sbp> AndyS: I doubt the rationale for including them in SQL engines is all that strong since they're not as extensible. you can use URIs, and you'd have to make major modifications to the SQL engine itself

20:49:14 <timbl> cwm will run qith sql if you tell it in metadata which sql database to use for which properties. But it will do the builtins internally. Converting them to SQL would ne neat.

20:49:14 <sbp> s/you can use URIs/you can't use URIs/

20:49:22 <timbl> SOURCE?

20:50:30 <adrianw> Left field question -- can SPARQL handle recursion?

20:50:43 <AndyS> Proposal is SPARQL query is over a collection of graphs. SOURCE accesses the origins of the graphs but the whole thing is also an RDF merge of all the graphs

20:51:35 <JosD> is like @forAll <#S>, <#P>, <#O>. [] q:select {(<#O> <a.n3>) a q:Answer}; q:where {<a.n3>.log:semantics log:includes {<#S> <#P> <#O>}}.

20:51:46 <AndyS> adrianw - "no" with a touch of "yes" - can run over an inferencing graph that does (e.g.) subClassOf transitivity or any other rules. Not part of the language though (to date anyway)

20:51:56 <sbp> { <input.n3>.log:semantics [ log:includes { ?s ?p ?o } ] } => { ?s ?p ?o } .

20:52:05 <gk> centre-left field question: can CWM handle recursive rules now? It used not to (part of my reason for embarking on Swish)

20:52:25 <JosD> indeed sbp :-)

20:52:32 <sbp> ah, that reminds me

20:52:36 <sbp> yosi_: still around?

20:52:52 <sbp> or anyone that can help: why was the quantification level for ?vars changes?

20:52:56 <sbp> s/ges/ged/

20:53:10 <sbp> it was between CWM 0.7 and 0.8, IIRC

20:53:12 <yosi_> I'm around for another two minutes

20:53:28 <timbl> gk: recursive rules - I'm not esure whether it will do what you want. It will generate rules from rules.

20:53:28 <AndyS> Can have SOURCE ?var { (?s :p ?o) (?o :q ?z) } to match on one subgraph and bind its URI to ?var

20:53:39 <sbp> yosi_: hi. it made the example for the Boeing chap rather difficult

20:54:01 <timbl> But you can't test for something with a universally quantified vriable on the LHS.

20:54:11 <sbp> (http://lists.w3.org/Archives/Public/public-cwm-talk/2004OctDec/0003)

20:54:44 <timbl> sbp: changed?

20:54:56 <sbp> yosi_: I'd've thought that rules-writing-rules would be a good enough motivation for keeping univars quantified over the root formula of the document

20:55:05 <sbp> timbl: so the changelog said... lemme find it

20:55:22 <AndyS> SPARQL use case: query over collection of graphs. Tricks with DISTINCT to make it the same as RDF-merge for common impls

20:55:51 <gk> timbl, it was rules that apply themselves recursively down a list, generating a new list. My experiments, from some time ago, desctibed here: http://www.ninebynine.org/SWAD-E/Intro.html#CwmRulesExperiments

20:56:03 <timbl> There is a desirable of the language that if you take a big chunk of it and enclose it in {} that you get a formula with the semantci sof the thing you enclosed.

20:56:11 <timbl> (sbp)

20:56:34 <sbp> whoops, my utter mistake. it was changed for bNodes, not univars (in 0.8)

20:56:41 <AndyS> Hmm - [] q:select {(<#O> <a.n3>) a q:Answer} isn't going to give a set of answers - will have duplicates.

20:56:59 <sbp> timbl: ah, I see the argument

20:57:06 <timbl> suppose i write a rule and then say " {tmp = 20 } => { ...my rule ... }

20:57:32 <timbl> I want the rule to have the same semantcis when quoted, even when it uses _: or ?x

20:57:41 * sbp nods

20:57:58 <sbp> I think I concede on that

20:58:22 <timbl> ok :)

20:59:41 <timbl> So what it is wrong with using (for SOURCE)

21:00:18 * gk thinks it's time to rejoin my family. Thanks everyone.

21:00:24 <DanC> SOURCE can't be done in cwm? then I take back my endorsement in a recent WG meeting. I must have misunderstood.

21:00:24 <timbl> { ?var a TestDocument; log;includes { ?s ?p ?o } } => { ?var mentioned ?o }.

21:00:28 <sbp> thanks gk

21:00:38 <timbl> Bye gk

21:00:47 <timbl> Thanks everyone for participating BTW

21:01:11 <sbp> a pleasure as usual

21:01:55 <adrianw> Possible topic for a future meeting: convergence between Query and Rule languages -- example at http://www.reengineeringllc.com/demo_agents/DataModelling1.agent

21:02:07 <timbl> Is SOURCE not basically log:semantics ... log:includes ?

21:02:17 <AndyS> My last attempt: http://lists.w3.org/Archives/Public/public-rdf-dawg/2004OctDec/0287.html

21:02:29 <JosD> DanC: I have trouble to do @forAll <#A>,<#S>, <#P>, <#O>. [] q:select {(<#O> <a.n3>) a q:Answer}; q:where {<#A>.log:semantics log:includes {<#S> <#P> <#O>}}.

21:02:47 <sbp> adrianw: looks like a simple subProperty example; am I missing something?

21:02:55 <AndyS> I had trouble when LHS of log:semantics is a var.

21:02:56 <JosD> i.e. an unbound <#A>

21:03:07 <JosD> indeed Andy

21:03:23 <sbp> JosD: you could get that from command line using os:argv

21:03:35 <adrianw> Yes, simple example. But it is both a query and a rule language. YOu could run it to see...

21:03:42 <AndyS> Also, SPARQL queries RDF and "?var a TestDocument" has ?var as a graph, not an RDF term.

21:03:46 <JosD> not yet sbp :-(

21:04:02 * AndyS wishes RDF graphs were in RDF - closes the world

21:04:06 <sbp> not yet in Euler?

21:04:06 * JosD thanks sbp to remind..

21:04:12 <sbp> ah. yep

21:04:15 <DanC> move the variable from the LHS of log:semantics to the LHS of log:includes and it works, I think, AndyS

21:04:50 <adrianw> Or, for a more complicated QueryLang=RuleLang example, there's http://www.reengineeringllc.com/demo_agents/RDFQueryLangComparison1.agent

21:05:30 <sbp> ooh, nice definition

21:05:32 <sbp> "SOURCE ?var (?s ?p ?o) - When SOURCE ?var is given before a triple, the variable will be bound to all of the known *Graph Names* for that triple."

21:05:37 <sbp> - http://lists.w3.org/Archives/Public/public-rdf-dawg/2004OctDec/0287

21:06:03 <DanC> never mind what I said about moving the variable...

21:06:29 <DanC> ... but you have to give cwm a list of possible sources. I understood that to be part of dajobe's proposal (...0297)

21:06:33 <DanC> 287

21:06:33 <timbl> There is a tacit underlying assumption that the kb is the conjunction of a set of smaller KBs, to exactly one level.

21:06:34 * DanC can't type

21:06:48 * JosD is still thinking about DanC's move..

21:07:12 <sbp> what would you do for triples that've been inferred by previous rules?

21:07:20 <sbp> make up some phony source?

21:07:26 <AndyS> In dajobe's proposal the subgraph has two roles : subgraph and part of the RDF merge. Need to have two copies for cwm?

21:07:42 <DanC> yes, graphs resulting from inference need their own name, sbp; see my msg about this...

21:08:41 <DanC> ... on SOURCE and inference: http://lists.w3.org/Archives/Public/public-rdf-dawg/2004JulSep/0363.html

21:08:49 <sbp> thanks

21:08:56 <timbl> <> owl:imports <f1>, <f2>. <f1> a TestFIle. <f2> a Testfile. ?var a TestFile; log:includes { ?s ?p ?o } => ....

21:09:33 <sbp> "the all-singing cat presentation tool": heh

21:10:23 <timbl> Copies? Well, cwm is explicitly aware of documents and kbs apart from the main one which is just a graph. All the others are just graphs too. So you get what you want by being explicit about it. But you don't complicate the KB to me more than an RDF graph.

21:10:25 <AndyS> Is owl:imports a builtin then?

21:10:29 <adrianw> what is a cat tool?

21:10:44 <DanC> NAME

21:10:44 <DanC> cat - concatenate files and print on the standard output

21:11:06 <DanC> i.e. unix cat.

21:11:12 <timbl> You can say ( ?kb1 ?kb2 ) log:conjuncton ?kb3 to merge explicitly.

21:11:14 <adrianw> thanks

21:11:36 <timbl> owl:imports is handled by cwm only if you set it into mode=i

21:11:42 <timbl> --more=i

21:11:47 <timbl> --mode=i

21:11:51 <sbp> is the #_formula hack still being used?

21:12:08 <timbl> no

21:12:09 * DanC wonders if --mode=i is excericsed by the test suite...

21:12:22 * timbl wonders whether --mode=p is

21:12:34 * timbl wonders whether --mode=rse is ... i think so

21:12:45 <timbl> but not sure

21:12:54 <AndyS> So, if the N3QlL query can run over the log:conjucntion of all graphs, and not in the input graph, it might work. Need some examples.

21:13:46 <AndyS> In 0287, I had two copies.

21:13:46 <sbp> I wrote a semi-proposal once to use IFPs in a URI scheme, so you'd have something like <urn:ifp:${IFP_URI}:${IFP_VALUE}> - perhaps you could do something similar to that for formulae URIs. then again, log:semantics isn't an IFP. hmm... but if it were acting on a *representation* it would be

21:13:59 <sbp> and I suppose you'd want its inverse too. urmph

21:14:13 <timbl> You can say ( ?kb1 ?kb2 ) log:conjuncton ?kb3. ?kb3 log:includes { today a SunnyDay}. ?kb1 log:includes { tomorrow a SunnyDay }

21:14:15 <DanC> online.tests: test:arguments "--mode=rs dbork/defdoc1.n3 --think".

21:14:34 <DanC> but I don't see tests for mode=i ; so it's not yet supported

21:15:34 <sbp> log:semantics is kinda confusing in that sense. documents change over time, or even vary according to language etc. Can one change the Accept headers that CWM sends out within a query?

21:15:34 <DanC> actually, log:semantics is an IFP. cwm assumes resources don't change once they've been observed

21:15:51 <sbp> it does? ouch

21:15:52 <timbl> FP

21:15:58 <AndyS> So would need to make ?kb3 the target of all patterns? ie.e. for "WHERE (?x ?y ?z)"

21:16:16 <timbl> It evaluates a web fetch just once in a run.

21:16:48 <timbl> There is the assumption that all cwm processing will take a negligible time compared to the change of the web ;-)

21:16:52 <timbl> Andy: yes.

21:16:53 <AndyS> 3Store has copies of same changing URL at different times.

21:17:09 <sbp> timbl: imagine a document that reflects your HTTP request but in RDF/XML

21:17:41 <timbl> 3store internalizes the HTTP experience as triples?

21:17:49 <JosD> is what we also do, take a snapshot of the world.. once in a run

21:17:55 <DanC> timbl and I have talked about refinements of log:semantics that take, for example, a message (representing a timepoint of sorts) as another arg. but his claim that it isn't needed most of the tim has held, so far.

21:18:05 <sbp> hmm, actually you don't have to send anything that changes. better example would be a counter, or something that increments: the current time in RDF, e.g.

21:18:15 <AndyS> 3Store is forever gathering graphs up and storing them - its a persistent store

21:18:41 <timbl> But deos it store the HTTP headers and what it asked fro and the datetime?

21:18:59 <AndyS> I think it stores some details in a system graph.

21:19:07 <timbl> So it can be objective about what it knows and why, at the system level?

21:19:25 <timbl> I wondered about providing access to the metalevel.

21:19:28 <sbp> DanC: yeah, I can sympathise with that. the only place that this has really come up for me is in testing builtins that get evaluated multiple times for some reason. I was doing a GET on an incrementing counter once; this must've been before CWM cached its Web results because I recall it jumping up multiple times

21:19:29 <AndyS> RDFStore might store same in side the graph it read in (merged in, that is)

21:20:00 <AndyS> For exact details, you'd need to ask the owners

21:20:33 <timbl> yes. I'd like to find out whether one could build a cwm on a 3store. I think probably.

21:20:50 <sbp> metalevel: reminds me: what is 'META_NS_URI = "http://www.w3.org/2000/10/swap/meta#"' for, in llyn.py?

21:21:07 * timbl has to go now.

21:21:24 <sbp> foiled again!

21:21:24 * AndyS going soon

21:21:34 <timbl> That was for the URI of such a meta formula which records cwm's personal experience.

21:21:46 <sbp> ah, hmm

21:21:58 <sbp> doesn't seem it was ever implemented though?

21:22:28 * JosD also taking a bit of family time - thanks a lot for the meeting, was good

21:22:42 * sbp waves to timbl, AndyS, and JosD

21:22:53 <timbl> implemented ... I think it is where the log:contents cache is kept.

21:23:03 <sbp> thanks

21:23:08 <DanC> there's also an 'experience' variable somewhere

21:23:09 * timbl waves .. thanks everyone

21:23:15 <DanC> not sure if it's the same thing

21:23:22 <DanC> yes, thanks all. enjoyed it.

21:23:56 <DanC> DanC has changed the topic to: Semantic Web hack-n-chat http://rdfig.xmlhack.com/ logs http://www.ilrt.bris.ac.uk/discovery/chatlogs/rdfig/

21:24:18 * AndyS waves goodbye

21:25:00 <adrianw> Lots to think about. Thanks to all and bye.

21:28:41 <DanC> A:[http://www.timeanddate.com/worldclock/fixedtime.html?month=12&day=13&year=2004&hour=16&min=30&sec=0&p1=0|2004-12-13T16:30Z] was proposed and 2nded

21:28:41 <dc_rdfig> Added comment A12.