Session Start: Fri Sep 27 15:22:51 2002 [15:22] *** Now talking in #rdfcore [15:22] *** Topic is 'I already told Zakim.' [15:23] *** em has joined #rdfcore [15:23] Zakim, Jim_Hendler is DaveB [15:23] +DaveB; got it [15:23] I might be joining a little late today. [15:23] *** em changes topic to 'rdfcore teleconference 2002-09-27' [15:23] *** jjc has joined #rdfcore [15:23] *** RRSAgent has joined #rdfcore [15:23] * RRSAgent is logging [15:23] Whjat's the meeting ID? [15:24] *** jang has joined #rdfcore [15:24] zakim, what's the passcode? [15:24] 7332 [15:24] the conference code is 7332, AaronSw [15:24] +??P2 [15:24] +??P6 [15:24] *** gk is now known as gk-scribe [15:24] Zakim, ??p2 is PatrickS [15:24] +PatrickS; got it [15:24] +AaronSw [15:24] zakim, who's here? [15:24] On the phone I see ??P0, ??P1, DaveB, Bwm, PatrickS, ??P6, AaronSw [15:24] *** DanC has joined #rdfcore [15:24] On IRC I see jang, RRSAgent, jjc, em, gk-scribe, DaveB, bwm, Zakim, AaronSw, logger [15:24] zakim, ??p6 is GK [15:24] +GK; got it [15:24] zakim, who's talking? [15:25] +??P21 [15:25] AaronSw, listening for 10 seconds I heard sound from the following: ??P0 (21%), Bwm (19%) [15:25] +??P22 [15:25] zakim, ??p21 is JosD [15:25] +JosD; got it [15:25] Zakim, ??p21 is jos [15:25] sorry, bwm, I do not recognize a party named '??p21' [15:25] zakim, ??p22 [15:25] I don't understand '??p22', AaronSw [15:25] Zakim, ??P22 is JJC. [15:25] +JJC.; got it [15:25] *** JosD has joined #rdfcore [15:25] Zakim, who's talking? [15:25] zakim, ??p22 is ILRT [15:25] sorry, AaronSw, I do not recognize a party named '??p22' [15:25] zakim, ??p21 is jos [15:25] sorry, bwm, I do not recognize a party named '??p21' [15:25] jjc, listening for 10 seconds I heard sound from the following: ??P0 (45%), Bwm (34%), GK (5%), AaronSw (18%), JJC. (9%) [15:25] zakim, ??p0 is ILRT [15:25] +ILRT; got it [15:25] -ILRT [15:26] zakim, ILRT has JanG [15:26] sorry, AaronSw, I do not recognize a party named 'ILRT' [15:26] quiet now? [15:26] no [15:26] +EricM [15:26] ok, back in a tick [15:26] Zakim, who is on the phone? [15:26] On the phone I see ??P1, DaveB, Bwm, PatrickS, GK, AaronSw, JosD, JJC., EricM [15:26] + +1.650.965.aabb [15:26] +??P20 [15:26] zakim, ??P20 is JanG [15:26] +JanG; got it [15:26] zakim, 1.650 is Guha [15:26] sorry, AaronSw, I do not recognize a party named '1.650' [15:26] zakim, ??p1 is Steve [15:26] zakim, +1.650.965.aabb is guha [15:26] +Steve; got it [15:26] +Guha; got it [15:27] zakim, who is on the phone? [15:27] On the phone I see Steve, DaveB, Bwm, PatrickS, GK, AaronSw, JosD, JJC., EricM, Guha, JanG [15:27] +DanC [15:27] +Manola [15:27] *** Guha_ has joined #rdfcore [15:27] gosh, lots today [15:27] zakim, who is on the phone? [15:27] *** danb_lap has joined #rdfcore [15:27] On the phone I see Steve, DaveB, Bwm, PatrickS, GK, AaronSw, JosD, JJC., EricM, Guha, JanG, DanC, Manola [15:27] +??P26 [15:28] perhaps we should re-vote with all these people [15:28] zakim, ??p26 is Sergey [15:28] +Sergey; got it [15:28] * danb_lap on his way [15:28] zakim, pick a victim [15:28] * jang isn't here next week, sorry [15:28] Not knowing who is chairing or who scribed recently, I propose Bwm [15:28] heh [15:28] +Mike_Dean [15:28] zakim, pick a victim [15:28] Not knowing who is chairing or who scribed recently, I propose Mike_Dean [15:29] zakim, pick a victim [15:29] Not knowing who is chairing or who scribed recently, I propose GK [15:29] zakim, pick a victim [15:29] Not knowing who is chairing or who scribed recently, I propose DanC [15:29] yeah, people goingto/webont next weekend? might affect attendance? [15:29] zakim, pick a victim [15:29] Not knowing who is chairing or who scribed recently, I propose Steve [15:30] Scribe next week: eric [15:30] zakim, who is on the phone? [15:30] On the phone I see Steve, DaveB, Bwm, PatrickS, GK, AaronSw, JosD, JJC., EricM, Guha, JanG, DanC, Manola, Sergey, Mike_Dean [15:30] *** mdean has joined #rdfcore [15:30] (that's DanC, not danbri) [15:30] danbri? [15:30] danbri said he'd dial in late [15:30] bwm said "Zakim sees... Dan ..." [15:30] Regrets: nine [15:31] Regrets: none [15:31] Agenda: OK [15:31] Next telecon: 20021004 [15:31] Last meeting minutes: [15:31] zakim, mute jang [15:31] JanG should now be muted [15:31] Clarify, there *was* a decision last week [15:31] zakim, who's talking? [15:31] AaronSw, listening for 10 seconds I heard sound from the following: PatrickS (4%), Bwm (62%), GK (25%), DanC (4%) [15:31] Status completed actions: [15:32] OK [15:32] zakim, temporarily mute gk [15:32] GK should now be muted [15:32] New HLink draft: [15:32] zakim, unmute gk [15:32] GK should no longer be muted [15:32] GK should now be unmuted again [15:32] Can anyone review this? [15:32] * DaveB suggests AaronSw - you like html, right? [15:32] gk-scribe, you've got a lot of noise on your line [15:32] No volunteers at this time. [15:32] 8. semantics untyped literals [15:33] eric's message - http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Sep/0313.html [15:33] All have seen Eric's message [15:33] Summary: ask chairs to reopen decision for more discussion [15:33] * danb_lap phone trouble, retrying [15:33] (1) concern with split decision, where group is responsibe to convince others... [15:34] (2) Schedule -- WG is overrunning -- decision is likely to extend CR period [15:35] If we get back to functional requirements, we may be able to find stronger consensus. [15:35] zakim, unmute jang [15:35] JanG should no longer be muted [15:35] JJC: second proposal; also JosD, Sergey [15:35] GK, mind if I mute your line? [15:36] BWM: needs new information to reopen. Is reopening within W3C process, if consensus of WG? [15:36] q+ [15:36] * Zakim sees Jang on the speaker queue [15:36] * gk-scribe danbri: any input yet? [15:36] dnabri, do you have a problem with reopening the issue [15:36] Zakim, who's talking? [15:36] zakim, mute gk-scribe [15:36] sorry, em, I do not see a party named 'gk-scribe' [15:36] jjc, listening for 10 seconds I heard sound from the following: Bwm (12%), GK (89%), AaronSw (30%), JJC. (25%), DanC (24%) [15:36] zakim, mute gk [15:36] GK should now be muted [15:37] * gk-scribe I didn't hear my noise ?-) [15:37] Zakim, mute Jjc. [15:37] DECISION: reopened on basis activity lead has so requested [15:37] 'the conference is full, no more parties can be added at this time'! [15:37] JJC. should now be muted [15:37] heh! [15:37] I can't dial in. [15:37] we are full [15:37] want me to drop out, danbri? [15:37] I support re-opening the issue per EricM's msg, fwiw. [15:37] nope. [15:38] thank you [15:38] I also support it [15:38] is there anyway to make the telecon larger? if not, i can drop out... [15:38] * danb_lap investigating [15:39] * gk-scribe eric looking to increase size of conf [15:39] eric: I will take care of this here in a second [15:39] I'll join soon [15:39] +DanBri [15:40] wohoo! :) [15:40] hm, i didn't hear danbri's join beep [15:40] * danb_lap wonders what he missed [15:41] q+ to ask about process effcet on other WG [15:41] * Zakim sees Jang, Gk on the speaker queue [15:41] zakim, who is on the phone? [15:41] On the phone I see Steve, DaveB, Bwm, PatrickS, GK (muted), AaronSw, JosD, JJC. (muted), EricM, Guha, JanG, DanC, Manola, Sergey, Mike_Dean, DanBri [15:42] * gk-scribe which document -- I missed something [15:42] the document is "pros and cons and look at requirement s etc" [15:42] the requirements doc? [15:42] ie, summarise the arguments rather than reiterate endlessly by email [15:43] Brian will assemble requirements doc [15:43] Steve: RDF is schizo -- trying to serve 2 communities -- AI and metadata -- and 3rd? [15:44] Steve own requirements, driven by metadata [15:44] "Inline" literals are required, don't care about model theory [15:45] I think that if Steve cares about the results he will get from APIs and queries, he should care about the model theory. [15:45] Uses for writing RDF directly, or produced by tools? The latter. But elegance / verbosity *is* an issue for bootstrapping [15:45] Later - less of problem. [15:45] I think steve might see it much as th XMP guys -- i.e. as a storage format [15:45] -Steve [15:46] DaveB: from metadata PoV. Asked by DC to talk to RDF issues. Inline and datatyped literals are required. Tidy/untidy not on radar. Mostly, need things finished. [15:46] Tidy is sufficient. [15:46] patrick: [15:47] From metadata perspective, consumed by machines not humans, controlling processes, etc. Need clarity of KRep aspects. Need to be able to generalize at higher levels without redundant re-specification (e.g. datatype on every instance) [15:48] In current data, inline literals *are* used. [15:48] that's an odd use of the term backwards-compatibility [15:48] we should retroactively reinterpret broken RDF correctly? [15:48] GK: [15:48] zakim, unmute gk [15:48] GK should no longer be muted [15:49] for myself, [15:49] i see no real difference in terms of functionality [15:49] speaking cor ccpp [15:49] it's clearly advantageous if the design expectations of ccpp could be satisfied [15:49] [can't hear this] [15:50] ccpp schema: change to tidy literals would need change to ccpp schema [15:50] don't think that's fatal [15:50] but there's a wg in the final throws of trying to get crec out [15:51] I'm unsure of the process in that case [15:51] finally, [15:51] hmm... graham, do you really need the CC/PP *schema* to not change? or just the instances? [15:51] there's an issue of consistency of treamtment between literal values and values represented as nodes [15:51] gk: don't want instances to change [15:51] it would be preferable if the schema doesn't have to change. [15:52] there's an issue of consistency of treamtment between literal values and values represented as nodes [15:52] zakim, mute gk [15:52] GK should now be muted [15:52] (that's a desirable if not a requirement) [15:52] zakim, who is on the phone? [15:52] On the phone I see DaveB, Bwm, PatrickS, GK (muted), AaronSw, JosD, JJC. (muted), EricM, Guha, JanG, DanC, Manola, Sergey, Mike_Dean, DanBri [15:52] Aaron: [15:52] need simplicity. Don't add more things to explain. [15:53] No real need for datatyping [15:53] Jos: [15:53] Requirement - getting at meaningful mechanized derivations [15:53] monoton, sound, not completeness [15:53] also: don't add more things to implement [15:54] things should be optional to those who don't need them [15:54] Jeremy: [15:54] Zakim, unmute jjc. [15:54] JJC. should no longer be muted [15:55] No clear *requirements* -- work currently focused in infrastructure issues [15:55] [jjc puts foot in mouth. details shall be lost to posterity :-) ] [15:56] Eric: [15:56] Zakim, mute Jjc. [15:56] JJC. should now be muted [15:56] Simplicity, listen to customers, interop across APIs [15:56] Guha: [15:56] yeah, simple [15:57] k.i.s.s. [15:57] (from guha) [15:57] API tends to work on data model, prefer simple model easy to explain, prefer tidy [15:57] Jang: [15:57] jang is 18 [15:57] lol [15:57] guha, you don't need any datatyping functionality at all? [15:58] Like a consistent story; prefer consistent simple to complete; implementation simplicity; would like to see demo that untidy is straightforwardly implementable [15:58] DanC: [15:58] (1) interoperability between APIs and query systems, with a clear logical justification behind it (e.g. a model theory) [15:58] I don't see untidy impls appearing soon; since IMHO existing impls are tidy or abivilant [15:58] (2) buy-in and understanding from DAML/WebOnt, CC/PP, RSS, and Dublin Core. [15:58] ... and Adobe/XMP, and DMOZ, and mozilla. And Jena. [15:58] -AaronSw [15:58] (g1) the ability to put, in a schema, constraints on the *lexical form* of values [15:59] +AaronSw [15:59] (g2) an endorsement of the practice of using datatype properties, i.e. [ dc:date "2002-09-23"]. [15:59] g1 == gravey-1; g2 == gravey2 [15:59] gravy/goal. [16:00] Frank: [16:00] zakim, who is here? [16:00] On the phone I see DaveB, Bwm, PatrickS, GK (muted), JosD, JJC. (muted), EricM, Guha, JanG, DanC, Manola, Sergey, Mike_Dean, DanBri, AaronSw [16:00] On IRC I see mdean, danb_lap, Guha_, JosD, DanC, jang, RRSAgent, jjc, em, gk-scribe, DaveB, bwm, Zakim, AaronSw, logger [16:00] wanto to be able to reflect datatyped information from databases and other typed sources like XML schema [16:00] oops.. s/dc:date/xsdt:date/ [16:01] support and annotate type info as metadata [16:01] support SWeb layers above RDF [16:01] type inference on a global as well as local basis [16:01] want to be able to infer type information globally and locally [16:01] means to deal with legacy data/code [16:02] Judge in terms of simple vs complex explanations. tidy/untidy no difference. Need to be clearer about what we mean. [16:03] er... the 10 test case isn't clear enough? [16:03] Sergey: [16:03] * DaveB also supports sergey's email, frankm's comments, aaron's comments [16:03] 1. simplicity, understandability for developers [16:03] 2. correctness for modelling [16:04] 3. backward compatibility (existing apps, specs, APIs) [16:04] MikeD: [16:04] 1. support for global typing [16:05] 2. avoid requiring apps to maintain lookup of ??? [literal?] values [16:05] in defense of mike, our implementation of tidy literals does have such a hash table. [16:05] 3. ability to merge duplicate statements (jenny age "10") in different documents [self-entailment?] [16:06] ... idependently on range constrains on "age" [16:06] 4. minimizing nodes and arcs -- avoiding extra bNodes Jenny age "10" prefer just one arc. [16:07] 5. support XMLschema built-in datatypes [16:07] DanBri: [16:07] requirement: something well understood, and simple to implement, soon. [16:07] datatyping: I want my database to be able to answer questions based on datatypes (before, after, during re dates; lessthan, gtr than) so need some conventions (almost doesn't matter which) to support this [16:08] As a consumer of RDF content, I want my db to not have to keep track of each different occurance of some literal. As a producer and consumer I need datatyped literals. [16:08] As a consumer of content... want database to NOT keep track of different occurrences of a literal. [16:08] +1 on all danbri says [16:09] [ I'm building query and storage systems for 'metadata' and data aggregation apps. For non-datatyped literals, tidyness works for me, and untidyness rings alarm bells. My main concern currently is that we get done soon. While my technical judgement is pro-tidy, I fear a decision to go untidy will mire us in months more discussion and debate. Tidy can seem inelegant/graceless, but is in my implementation experience both simple and deployable. I can't say the s [16:09] Tidy may seem inelegant and graceless, but is also simple and deployable. [16:09] ACTION, bwm: summarize requirements into next iteration of rationale document [16:09] Zakim, unmute jjc. [16:09] JJC. should no longer be muted [16:09] gk-scribe, just to note I pretty much read out from text I pasted above; shouldn't be missing anything important [16:09] Item 10 -- abstract syntax for untidy literals [16:10] Do untidy literals have a "sysID"? [16:11] JJC: position - No untidy literal is ever the object of two statements. But reification is then odd. [16:11] ... i.e. every occurrence of a literal is a distinct node -- distinct by triple/position it occurs in [16:12] "10" . [16:12] "10" . [16:12] one triple or two in this case, jjc? [16:12] ... in N-triples, distinguish by (e.g.) line number. [16:12] holy cow... this ["each literal occurs in exactly one triple"] cuts off the possibility of an RDF/XML syntax that allows literals as subject. phpht. [16:12] yeah, sort of [16:13] e.g. for Jena, do lexical compare without type info, but value if type available, but this is non-mon. [16:14] Jenny age "10" . Film title "10" . does not entail ... age _:x ... title _:x [16:15] that (hypothetical) behaviour by Jena is exactly the sort of thing that's counter to my #1 requirement for datatypes: that the behaviour of APIs have a good logical basis. Jeremy just said it's nonmon. [16:16] Possible future decision to have sysID on literals -- this is implementation issue, not MT [16:17] Guha, don't understand this -- breaking existing large implementations (Mozilla, ...) [16:18] Brian: clarifies that the point of this is to be clear about what untidy means when making decision. [16:19] q+ to ask for meeting extension to get some decisions :) [16:19] Guha: Difference between "could be differnt" and "have to be different". [16:19] * Zakim sees Jang, Gk, DaveB on the speaker queue [16:19] the implementation Guha mentioned was XUL, part of the mozilla project. [16:19] q- [16:19] q- [16:19] * Zakim sees Gk, DaveB on the speaker queue [16:19] * Zakim sees DaveB on the speaker queue [16:20] -AaronSw [16:20] then daveb [16:20] I consider guha's point in part considering XUL, Mozilla an important 'customer' and any decision we make re tidy/untidy should take the impact of this customer into account... [16:20] yup [16:20] and breaking the largest customers implementation should not be taken lightly [16:21] * DanC q+ to make the point that the behavour of APIs is how end-users will understand our spec; in particular, our model theory [16:21] * Zakim sees DaveB, DanC on the speaker queue [16:22] q+ [16:22] * Zakim sees DaveB, DanC, Jang on the speaker queue [16:22] ack daveb [16:22] DaveB, you wanted to ask for meeting extension to get some decisions :) [16:22] * Zakim sees DanC, Jang on the speaker queue [16:22] DaveB: Proposed to extend meeting, to make decision on sysIDs. Not granted. [16:22] *** Guha_ has quit IRC (ChatZilla 0.8.7 [Mozilla rv:1.0.0/20020530]) [16:22] ack jan [16:22] * Zakim sees DanC on the speaker queue [16:22] bah, no more decisions this week [16:23] JanG: how fast to get untidy implementation to see how it works? [16:23] ... ergo no docs change, no test changes, no code changes [16:23] JosD: done as experiment -- is a nightmare. [16:23] Jos, if you could mail the results of your untidy experiemnt, that would be nifty. [16:24] Sergey: hard to assess with 24000 lines of code that assume tidiness. [16:24] q+ to comment on mozilla api [16:24] * Zakim sees DanC, Danb on the speaker queue [16:24] summary: API changes are easy, application changes are hard [16:24] ... Do it in the API is one thing (easy?), do it in the calling application is harder [16:24] q- [16:24] * Zakim sees DanC on the speaker queue [16:24] * DanC q- [16:24] * Zakim sees no one on the speaker queue [16:24] well DanC the whole point is that I got nothing but "No Proof Founds" so there is not much to explain further [16:25] -Guha [16:25] ACTION, Sergey, write up what he just said [16:25] -JJC. [16:25] --------meeting closed------ [16:25] yes, well, showing all the "no proof found" stuff is likely to help folks understand. [16:26] -DanC [16:26] ...I tried to fix by all kinds of "extra wires" and that was the nightmare [16:26] again, very useful implementation experience to share. [16:27] zakim, who is on the phone? [16:27] On the phone I see DaveB, Bwm, PatrickS, GK (muted), JosD, EricM, JanG, Manola, Sergey, Mike_Dean, DanBri [16:28] ... the only thing that is reasonable is make it <#foo> <#bar> (<#bar> "abc") but only for <#foo> a rdfs:Datatype . [16:28] JosD: what if there were a common superproperty for all rdf properties [16:28] -JosD [16:29] oops, sorry Brian I had to hang up, but w.r.t. your questio [16:30] * danb_lap has to head off; cu [16:30] -Mike_Dean [16:30] *** danb_lap has left #rdfcore (danb_lap) [16:31] ... I don't think that is a problem at all i.e. it is not breaking inferencing, but I think further and put something on the list [16:32] it means that "foo" entails "foo" for all , and hence all must have some datatype [16:32] zakim, unmute gk [16:32] GK should no longer be muted [16:36] OK I see your point but is a Datatype and so are it's superproperties; Ireally think this is no problem [16:36] -DanBri [16:37] * DaveB drops out, bye chaps [16:37] -DaveB [16:37] xsd:integer"10" =? xsd:integer"010" [16:38] ... at least according to our tests using http://www.agfa.com/w3c/euler/rdfs-rules [16:49] *** bwm_ has joined #rdfcore [16:54] -EricM [16:55] -GK [16:55] *** bwm has quit IRC (Ping timeout)