07:03:39 RRSAgent has joined #ldp 07:03:39 logging to http://www.w3.org/2013/06/19-ldp-irc 07:03:41 RRSAgent, make logs public 07:03:41 Zakim has joined #ldp 07:03:43 Zakim, this will be LDP 07:03:43 ok, trackbot; I see SW_LDP(F2F)2:30AM scheduled to start 33 minutes ago 07:03:44 Meeting: Linked Data Platform (LDP) Working Group Teleconference 07:03:44 Date: 19 June 2013 07:04:30 SteveS has joined #ldp 07:08:59 nmihindu has joined #ldp 07:10:43 agenda: http://www.w3.org/2012/ldp/wiki/F2F3#Day_2_-_Wednesday_June_19 07:11:01 scribe: nmihindu 07:11:13 chair: Arnaud 07:14:59 Arnaud: we will start discussing the issues later so that people connecting from the US can also participate 07:15:10 Topic: Access Control WG Note - steps towards FWD 07:15:22 Arnaud: what is the status ? 07:15:24 where is the note? 07:15:46 http://www.w3.org/2012/ldp/wiki/AccessControl 07:16:52 asok: we have the note, we need to get it reviewed by the WG 07:17:53 Arnaud: we can give an action item for getting this note reviewed 07:18:11 ashok: Ted was interested in reviewing the note 07:18:55 mesteban__: what is the deadline for reviewing the note ? 07:19:24 ashok: normally it is two weeks, but it can take more 07:19:52 JohnArwe has joined #ldp 07:20:12 Arnaud: our priority is the spec, other things come second 07:20:35 Arnaud: Ted and mesteban__ can review the document 07:20:35 action: mesteban to review and comment the WG Access Control draft 07:20:35 Created ACTION-76 - Review and comment the WG Access Control draft [on Miguel Esteban Gutiérrez - due 2013-06-26]. 07:21:05 action: Ted to review and comment the WG Access Control draft 07:21:05 Created ACTION-77 - Review and comment the WG Access Control draft [on Ted Thibodeau - due 2013-06-26]. 07:23:08 Arnaud: deadline for reviewing these document will be 1 month 07:23:18 ... anything more to discuss on this ? 07:23:24 AndyS has joined #ldp 07:23:48 mielvds has joined #ldp 07:23:52 Topic: Deployment Guide - steps towards FWD 07:23:58 where is the deployment guide? 07:24:15 http://www.w3.org/2012/ldp/wiki/Deployment_Guide 07:25:21 krp has joined #ldp 07:26:06 bblfish: what is the difference between the deployment guide and best practices ? 07:26:39 Arnaud: we need to find an editor for the deployment guide ? 07:27:11 Ashok: what is the difference between the primer and the deployment guide ? 07:28:05 Arnaud: Primer is the primarily for users and the deployment guide is mainly for implimenters 07:28:13 SW_LDP(F2F)2:30AM has now started 07:28:20 +m 07:28:33 ... the goals and the audience are different 07:29:29 ... we moved some stuff from the spec to the deployment guide 07:29:54 Ashok: why did we move datatypes from the spec to deployment guide ? 07:30:04 q+ 07:30:43 ack nmihindu 07:30:45 q? 07:33:46 cody has joined #ldp 07:34:28 Arnaud: there were things in the spec that would better fit into the deployment guide, so we moved them from the spec to the deployment guide 07:35:17 ... spec defines the conformance and and the deployment guide shows best practices 07:36:35 cody: I can help to do the editorial stuff and organizing it better 07:37:45 cody: is there a deadline for this ? 07:37:59 Ashok has joined #ldp 07:38:12 Arnaud: we don't have a specific deadline as per now 07:38:23 Re. RDF datatypes, here is a useful document: http://www.w3.org/TR/swbp-xsch-datatypes/ 07:38:28 rgarcia has joined #ldp 07:39:07 Arnaud: we can make cody the primary editor and nmihindu to help 07:39:14 Confirmed: I will be the primary editor for Deployment Guide with Nandana as assist. 07:40:13 Arnaud: any more issues discuss on this topic ? 07:40:35 SteveS: the name deployment guide is confusing 07:40:54 Arnaud: we can change the name now, there are few proposals in the wiki 07:41:28 cody: deployment in software is very different from what we have in the document 07:41:58 cody: LDP best practise and guidelines ? 07:41:58 suggested title: LDP Best Practices 07:42:12 suggested title: LDP Best Practices and guidelines 07:42:43 I'm good with: LDP Best Practices and Guidelines 07:43:27 bblfish: deployment is more about publishing your data 07:44:24 Arnaud: LDP best practices is generic enough to cover everything we have in the document 07:45:56 Proposal: A: LDP Best Practices B: LDP Best Practices and Guidelines 07:46:05 PROPOSED: Change title of deployment guide to "LDP Best Practices and Guidelines" 07:46:07 B 07:46:11 A 07:46:23 A 07:46:29 A 07:46:40 No B 07:46:45 I can perfectly live with A 07:46:46 Its B! 07:46:57 A 07:47:08 B 07:47:11 B, can live with A 07:47:23 B 07:47:48 Ashok_ has joined #ldp 07:47:51 I can live with A, say +0.51 B, +0.49 A 07:48:48 rgarcia: let's let cody decide 07:49:05 Resolved: Change title of deployment guide to "LDP Best Practices and Guidelines" 07:50:08 Topic: Test Suite & Validator - steps to completion 07:50:46 Arnaud: it would have been better if ericP was here 07:50:50 ericP you awake? 07:51:17 rgarcia: I will give an overview and the next steps 07:52:26 https://dvcs.w3.org/hg/ldpwg/raw-file/default/Test%20Cases/LDP%20Test%20Cases.html 07:53:42 rgarcia: we have defined test cases for core LDP features - all the MUSTs in the spec 07:54:07 rgarcia: explaining the current status of the document 07:55:34 ... the document defines the test cases and result and they are linked for traceability 07:56:47 ... the tests can be run by a software automatically or manually and the results can be submitted 07:57:29 ... In the spec, we have both testable requirements and non-testable requirements 07:57:49 ... it is better to have testable requirements 07:59:24 ... there are some ambiguities in the spec, we need to remove them to make all the requirements are testable 08:00:45 Arnaud: this for implementaters to test their implementations 08:01:03 ... not a test harness 08:02:16 mesteban__: we already discussed this, there are a lot issues testing application specific LDP implementations 08:05:17 bblfish: test cases can help to find the problematic areas of the spec 08:05:53 rgarcia: at the moment, we cover all the MUSTs but not different compliance levels etc. 08:06:24 Arnaud: how much are we missing ? Paging, Sorting ? 08:07:03 rgarcia: we are missing the SHOULDs 08:07:28 Arnaud: Is anybody using the test suite already ? 08:08:03 ... it would be interesting to use it and provide feedback 08:08:41 mielvds has joined #ldp 08:08:53 cody_ has joined #ldp 08:09:09 mesteban__: we can not have a test harness for application specific LDP implementations 08:09:30 Arnaud: we have at least have one test harness for vanilla implementations 08:10:07 mesteban__: if the developers can provide their data, we can provide a SPI for executing the tests 08:10:57 Arnaud: we need at least two implementations compliant with the spec 08:13:06 Arnaud: we need to find people responsible for coming up with a harness and generate the report 08:13:17 Alex Bertails had promised to work on the implementation for the test harness ( using Banana RDF possibly ) 08:14:41 rgarcia: the current tests can be run manually and provide the results to us 08:16:03 cody: we need to define a standard format for reporting the results 08:16:31 rgarcia: it is already defined in the document 08:17:06 bblfish: I can volunteer to provide a test harness 08:17:16 with the help of alex bertails. 08:17:18 :-) 08:17:22 ( but will do it ) 08:17:24 roger has joined #ldp 08:17:30 +q 08:17:51 ack roger 08:18:15 I can contribute on @bblfish his github repo 08:18:20 Arnaud: we can provide a harness for vanilla implementations and the application specific LDP servers can start from there and define their own 08:18:47 I'll post this to the group. Will use banana-ref https://github.com/w3c/banana-rdf 08:18:53 roger: other standards do interop fests, can we do something like that ? 08:19:45 Arnaud: we can do that, it is always helpful to improve interoperability and also find ambiguities in the spec 08:20:53 Arnaud: how we can improve the test suite to include conformance, affordances etc ? 08:21:48 ... we define the different conformance levels with names, we can make set of tests to cover specific conformance 08:22:54 rgarcia: it makes more sense to have separate them as modules, so they are not built on top each other but rather can be orthogonal 08:23:55 https://dvcs.w3.org/hg/ldpwg/raw-file/default/ldp.html#http-patch 08:23:59 bblfish: can you provide an example of specific spec sections which are not testable ? 08:25:51 rgarcia: we have you MAY implement feature X, and then SHOULD. It is better to say if you implement feature X, then you MUST DO these 08:25:51 jmvanel has joined #ldp 08:26:21 q? 08:26:22 q+ 08:27:32 ack bblfish 08:27:40 ... it is hard to check the pre-conditions at the moment, for example whether we can do PUT on a resource 08:29:15 Arnaud: do you think some of the SHOULDs must be changed to MUSTs ? 08:31:04 ... in a way, if you implement some feature (or a module), then you MUST and MUST NOT do several things 08:31:20 ... we can reduce the number of SHOULDs 08:34:28 JohnArwe: if we organize the spec like modules, it could be helpful to the implementors to only focus on specific modules they would like to implement 08:37:09 Coffee break !! 08:54:02 stevebattle3 has joined #ldp 08:54:35 rgarcia has joined #ldp 08:55:35 back from coffee 08:57:21 Zakim, who's on the phone? 08:57:21 On the phone I see m 08:57:44 i thought m died at the end of the last movie 08:58:42 +[GVoice] 08:58:48 Zakim, [GVoice] is me 08:58:48 +ericP; got it 08:59:12 SteveS has joined #ldp 09:00:25 scribe: rgarcia 09:00:30 mielvds has joined #ldp 09:00:55 topic: Test Suite 09:01:25 ericP: I may implement a test harness, but maybe not in time 09:01:49 ashok: Eric, did you implement something for RDB2RDF? 09:02:04 ericP: That case was much simpler 09:04:49 Arnaud: The point is that if someone writes the test harness for a generic LDP server other people can reuse it, even for the domain-specific LDP servers 09:04:53 mesteban___ has joined #ldp 09:05:03 nandana has joined #ldp 09:05:29 bblfish: I plan to implement a test harness 09:06:02 mesteban___ has joined #ldp 09:06:18 ericP: Alexandre said that he was going to implement something but would be quite specific, maybe not of value for others 09:06:56 Arnaud: Right now there is no one in charge of developing and maintaining a test harness 09:08:34 Arnaud: Back to discussing issues 09:08:52 topic: Modules / profiles / affordances 09:09:02 Arnaud: What do we need to have? 09:09:22 -> https://github.com/ericprud/SWObjects/blob/sparql11/tests/test_LDP.cpp#L550 generic triple store LDP test 09:09:28 JohnArwe: Why should we do that? 09:09:33 ericP I see the test harness for RDB2RDF at http://www.w3.org/2012/ldp/wiki/Testing#Example_Usage_within_W3C 09:10:19 JohnArwe: … one benefit is to make the specification more consumable: by users, test suite, etc. 09:11:04 SteveS, indeed. i wonder what purpose those links fill there. perhaps inspiration? 09:11:58 ericP yes, just "hey, look at some other stuff" and maybe someone could factor out something reusable (HTTP commands/response) 09:12:21 roger: so, there will be different types of servers; for example with and without pagination 09:12:36 JohnArwe: and the clients may also want to use certain features or not 09:12:37 SteveS, true, that's the concrete vision that i think we'd need. 09:13:44 Arnaud: Is the read-only profile covered by OPTIONS? 09:13:53 +Sandro 09:13:56 JohnArwe: Yes 09:14:14 Arnaud: what else are we not addressing? 09:14:20 Issue-32? 09:14:20 ISSUE-32 -- How can clients discover that a resource is an LDPR or LDPC, and what features are supported? -- open 09:14:20 http://www.w3.org/2012/ldp/track/issues/32 09:15:08 Issue-80? 09:15:08 ISSUE-80 -- How does a client know which POST requests create new resources -- open 09:15:08 http://www.w3.org/2012/ldp/track/issues/80 09:15:49 http://www.w3.org/2012/ldp/wiki/ISSUE-32 09:16:10 cody has joined #ldp 09:16:43 Ashok has joined #ldp 09:17:05 JohnArwe: Does a server support membership triples whose object is not an LDPR? 09:17:25 bblfish: That is related to how do you know what can be posted to a container 09:17:33 JohnArwe: That is related to ISSUE-80 09:18:56 stevebattle4 has joined #ldp 09:19:32 JohnArwe: 4.1.3 is more related to create that to other things such as PATCH 09:21:08 JohnArwe: There are not so many affordances without a discovery mechanism 09:23:27 +SteveBattle 09:23:32 Ashok: How can it be detected that the data sent to the server is not valid? 09:23:49 SteveS: According to the datatype 09:24:20 JohnArwe: 5.4.3 is again about ISSUE-80 09:24:46 … and 5.6.x is about recursive delete which is already closed 09:25:05 … so we are covering everything except those things related to ISSUE-80 09:25:14 topic: ISSUE 80 09:25:25 ISSUE-80? 09:25:25 ISSUE-80 -- How does a client know which POST requests create new resources -- open 09:25:25 http://www.w3.org/2012/ldp/track/issues/80 09:26:01 JohnArwe: There is already a proposal for solving the issue 09:26:45 q+ 09:26:52 … replicating the way it is used in PATCH for POST would be nice from the HTTP perspective 09:27:10 … but the semantics of POST are completely open 09:27:49 suggest something like a link header that says { <> memberType :someType . } e.g. { <> memberType :Bug 09:27:55 } 09:28:32 … for example, POST is frequently used for querying 09:28:37 bblfish -1, that is overkill for what we are talking about and different issue 09:29:36 … we can leave things open (as everyone else in the web) or include service documents that specify them 09:30:10 q? 09:31:00 ack bblfish 09:31:14 bblfish: this doesn't solve the problem deep enough 09:31:23 q+ 09:32:02 … in our case everything is turtle and we want to distinguish what can I post to a container 09:32:24 … the mime type is the wrong approach, we should make it declaratively 09:32:30 suggest something like a link header that says { <> memberType :someType . } e.g. { <> memberType :Bug } 09:33:24 q+ 09:33:26 q+ 09:33:40 thinks that should just be non-member-properties, not link headers…but this is beyond this issue 09:34:09 ack sandro 09:34:46 roger has joined #ldp 09:34:47 +q 09:35:14 ack steves 09:35:25 sandro: outside of LDPCs, it would be also for resources to know what can be posted 09:35:52 SteveS: Part of the data is RDF and other part is not (e.g., binaries) 09:36:36 … we don't need to put that information in the link header, since it is part of the resource 09:36:36 SteveS: we need mime types and member types. 09:37:11 … but this leads us maybe beyond LDP 1.0 (constraints, etc.) 09:37:11 q+ to say that re-using RDF types is not the right granularity for what a server will accept 09:37:46 ack john 09:38:16 JohnArwe: bblfish, how does your proposal define the semantics of the operation? 09:39:00 .. is that a constraint of a hint? 09:39:09 My suggested layering is: what content-types are accepted on POST, does POST support create, then (beyond LDP 1.0 I believe) IF it is RDF content, what Classes are allowed/expected 09:39:28 bblfish: You may have more that one link 09:39:46 q+ 09:40:16 … what John is proposing would be a closed world assumptiom 09:40:18 q- 09:40:24 ack ashok 09:40:30 ericP: more than that, a closed protocol assumption 09:40:54 q+ to ask what prob we need to solve in 1.0 09:40:59 Ashok: You specify the type, but what about the schema? 09:41:19 … you also want to specify the structure 09:41:22 yes ericP, see my previous post 09:41:52 -Sandro 09:42:13 bblfish: you can add multiple relations 09:42:19 JohnArwe: but it gets complicated soon 09:43:58 Ashok: it is only useful if you can specify the properties of a type 09:44:45 Arnaud: the point is at what level do we specify those restrictions 09:44:57 ack roger 09:45:10 Orthogonal background question for Sandro or EricP... in the process of drafting the sorting stuff I noticed that ReSpec's SPARQL-QUERY reference (normative to LDP) points to a document http://www.w3.org/TR/rdf-sparql-query/ which says at the top "go see 1.1" ... should we be changing that ref now? 09:46:03 roger: there will be properties on containers beyond the type of what the container contains 09:46:48 ...well need to be careful with "contains" word. to some, means "created by container", to others "in the container's membership" 09:47:31 I think what Roger did was say that implementations might distinguish between the types of members that exist, and the types of new members that create would accept. 09:47:37 bblfish: at the end OWL is about sets and this is our case 09:48:47 … my proposal allows adding later more specific things so clients can be later more advanced 09:49:01 ack steveb 09:49:26 stevebattle4: I like in principle bblfish's proposal and I don't think we should go beyond that 09:49:37 your volume is highly variable steveb 09:49:47 so 'things inside container' and 'ability to create new ones of those things inside container'. 09:49:55 JohnArwe, ReSpec stuff is handled by a secret band of maintainers who, when prodded, update the table associating short name to a spec name. then we have to do a pull to use that updated ReSpec. 09:49:56 ... just to summarise a bit John 09:49:57 … I don't know a general way that can be useful for validation 09:50:39 Arnaud: In September there will be a workshop on RDF validation 09:50:43 -> http://www.w3.org/2012/12/rdf-val/SOTA#shapes example of OSLC's language for describing valid input 09:50:58 rdf validation workshop = http://www.w3.org/blog/SW/2013/05/22/w3cs-rdf-validation-workshop-practical-assurances-for-quality-rdf-data/ 09:51:29 ack eric 09:51:29 ericP, you wanted to ask what prob we need to solve in 1.0 09:51:56 ericP: is it worth to talk about validation approaches? 09:52:10 … but for 1.0 we should not cover this 09:52:47 Arnaud: resource shapes is a vocabulary to describe the resources you manage 09:53:14 I also think this causes deep semantic conflicts. This implies OWL reasoning to accept subclasses with every post, which can be a fraction of RDF validation, which would be able to cover all semantics 09:53:32 … in an RDF document 09:54:32 Arnaud: Right now we don't have a complete solution for validation, right now we have media types but beyond that we don't have nothing stable 09:55:46 +Sandro 09:56:12 bblfish: at the end anything will be something that defines a set of documents 09:56:31 … so we can link to that something 09:56:52 q+ 09:57:30 s/anything will be something/the solution will be a language/ 09:57:33 ack steves 09:58:09 Henry, It may only directly constrain the set of RDF models, and only indirectly the set of documents. 09:58:30 SteveS: we can specify the media type, but we cannot force now the specification of types 09:58:32 seems to come down to a small number of questions: (1) do we have consensus to add something about this (at any level) to the existing corpus (if n, done, else) (2) do we have consensus to add something telling clients which media types are accepted for create (regardless of how specified) 09:59:14 Arnaud: Can we leave the type declaration for later? 09:59:35 … E.g., LDP 1.1 09:59:39 +1 to later 09:59:41 ...(if no, done, else) (3) do we have consensus to add that at the HTTP and/or RDF levels (if no, done, else which and then) (4) do we have consensus to add anything more? 10:00:15 SteveS: Types and constraints may conflic 10:00:22 s/conflic/conflict/ 10:00:48 q? 10:00:56 bblfish: But we would be promoting bad behaviour 10:01:01 q+ 10:01:16 Arnaud: But that's the reason we have the Best Practices document 10:01:35 +1 to discouraging misuse of media type in the best practice and guidelines 10:02:39 ack john 10:03:18 q+ to say that no tool is going to be able to do anything with an english definition of the type 10:03:29 I don't see how this encourages abuse of media type, people can do POST today and create media types….how does this encourage it? 10:03:43 JohnArwe: Reads aloud the questions written above 10:03:50 q+ 10:05:25 q- 10:06:34 sandro: are the use cases for ISSUE-80 compelling? 10:07:09 There is no current use-case in UC&R that covers this. 10:07:10 bblfish: it helps the test suite 10:07:56 Arnaud: is someone against dealing with the media type question? 10:08:31 ack bblfish 10:08:51 bblfish: beyond media types we would like to define sets of documents 10:09:24 Arnaud: we seem to have a consensus on media types but not on the RDF typing 10:11:28 Revised Proposal: define a new HTTP header Accept-Post whose value is a media type list to communicate which media types the server accepts when creating resources via HTTP POST 10:12:34 +1 10:12:35 +1 10:12:38 ...the change vs -80 is removal of -Create suffix 10:12:46 +1 10:12:46 +1 10:12:48 +1 10:12:51 0 - I'd prefer to define a new RDF property accep-post whose value is.... 10:12:59 +1 10:13:00 +1 10:13:04 +1 (though i'm concearned about the loss of -Create) 10:13:10 s/accep-post/accept-post/ 10:13:14 -1 10:13:18 +0 10:13:31 +q 10:14:00 0 10:14:45 ack nmihindu 10:16:02 nmihindu: going behind this proposal may have risk 10:19:11 q+ 10:19:25 s/behind/beyond/ 10:19:46 ack krp 10:20:14 q+ to ask whether we can drop this requirement and drop the header 10:20:47 ack eric 10:20:47 ericP, you wanted to ask whether we can drop this requirement and drop the header 10:21:21 krp: There are different levels of constraints and bblfish proposal could also be implemented with the media type approach 10:21:35 +1 to drop this altogether 10:22:22 -1 to dropping this 10:23:02 Ashok, what can we do with this header? 10:23:56 Eric, I worry that the server will get fiilled up with bad data unless we constrain what you can add where 10:24:14 ack sandro 10:24:20 +q 10:24:45 sandro: How are we going to define this header? 10:24:55 JohnArwe: We can just put it in our specification 10:25:06 Ashok, we still have that issue with the header. all the header does is say that some endpoint accepts post 10:25:27 … and requires approval from a domain expert 10:25:31 homework for Henry then is there an ontology for the mime types ? 10:25:50 Eric, I want it to say what type it accepts 10:25:50 q+ 10:25:54 also if there are problems with documents types as thought of this way. 10:26:04 ericP, I think it is more like the reverse of Accept header in a GET for client. But this time server letting the clients know what the server will accept. 10:26:43 sandro: This is unrelated to LDP 10:26:59 q- 10:27:12 … it is an important thing to have and could be addressed elsewhere 10:27:53 do we need a straw poll on whether or not to STOP AT media type? If anyone is going to -1 that, then we know that no matter which way we attempt to go we're looking at a formal objection, so at that point we might as well look for "near consensus" 10:28:03 ack roger 10:28:05 http://tools.ietf.org/html/rfc5988#section-6.2.1 is the process for adding new link relations 10:28:21 q+ 10:28:35 +Sandro.a 10:28:38 q+ 10:28:43 -Sandro 10:28:45 Ashok: it seems that we are just postponing difficult topics 10:29:37 q+ to say that losing this header isn't a great loss; that we aren't giving up on any use cases. 10:29:46 ack steveb 10:30:28 +q 10:30:34 ericP: Adding a new header variable should be the last resort 10:30:45 Yes - that's what I said 10:30:57 Sorry for the sound qaulity 10:31:06 s/qaulity/quality/ 10:31:10 s/ericP/stevebatrle4/ 10:31:16 ack ashok 10:31:27 s/stevebatrle4/stevebattle4/ 10:31:38 cody has joined #ldp 10:32:04 s/stevebatrle4/stevebattle4/ 10:32:05 ack eric 10:32:05 ericP, you wanted to say that losing this header isn't a great loss; that we aren't giving up on any use cases. 10:32:54 ericP: If we do not add the header we would be ruling some use cases 10:32:55 ack roger 10:33:09 s/would be/would not be/ 10:33:29 ericP, yes we are. We already had consensus this was a requirement, 20 minutes ago -- and if we hadn't, I'd have given some use cases. 10:34:57 roger: we need a strategy as a group regarding all these things 10:35:18 … and avoid ad-hoc solutions every time we face this kind of decisions 10:36:45 q+ 10:37:46 q+ 10:37:56 ack sandro 10:37:58 Roger: how do things function in your view when the request-uri=R and all the triples in the response (i.e. the state of the resource) have a subject URI of S ? 10:38:18 Document from 2002 http://www.w3.org/2001/tag/2002/01-uriMediaType-9 10:38:33 sandro: Tim thought of URLs for media types but at the end he concluded that shouldn't be done 10:39:02 ack rgarcia 10:39:44 sandro: IETF showed no interest in mapping media types to URIs, not W3C's place to do so. 10:41:56 sandro, i agreed to the requirement 'cause i wanted to make progress, but i'd like to see some use case that gets enabled by accept-post 10:42:47 <> memberType [ mime "image/*" ] . 10:42:52 cody has joined #ldp 10:44:28