13:54:27 <RRSAgent> RRSAgent has joined #ldp
13:54:27 <RRSAgent> logging to http://www.w3.org/2014/03/24-ldp-irc
13:54:29 <trackbot> RRSAgent, make logs public
13:54:31 <trackbot> Zakim, this will be LDP
13:54:31 <Zakim> ok, trackbot, I see SW_LDP()10:00AM already started
13:54:32 <trackbot> Meeting: Linked Data Platform (LDP) Working Group Teleconference
13:54:32 <trackbot> Date: 24 March 2014
13:54:33 <sandro> zakim, who is on the call?
13:54:33 <Zakim> On the phone I see [IPcaller]
13:54:47 <sandro> zakim, IPcaller is Sandro
13:54:47 <Zakim> +Sandro; got it
13:59:58 <Zakim> +Roger
14:00:12 <Zakim> -Roger
14:00:54 <Zakim> +Arnaud
14:01:37 <sergio> sergio has joined #ldp
14:01:39 <sergio> hi
14:01:44 <TallTed> TallTed has joined #ldp
14:01:56 <Zakim> +Roger
14:01:59 <Arnaud1> Arnaud1 has changed the topic to: LDP WG: http://www.w3.org/2012/ldp - next agenda: http://www.w3.org/2012/ldp/wiki/Meetings:Telecon2014.03.24l
14:02:21 <SteveS> SteveS has joined #ldp
14:02:41 <Ashok> Ashok has joined #ldp
14:02:44 <roger> roger has joined #ldp
14:02:50 <Zakim> + +43.660.274.aaaa
14:02:54 <Zakim> +[IBM]
14:03:05 <SteveS> Zakim, [IBM] is me
14:03:05 <Zakim> +SteveS; got it
14:03:11 <JohnArwe> JohnArwe has joined #ldp
14:03:16 <Zakim> +OpenLink_Software
14:03:17 <Zakim> +Ashok_Malhotra
14:03:38 <TallTed> Zakim, OpenLink_Software is temporarily me
14:03:38 <Zakim> +TallTed; got it
14:03:40 <TallTed> Zakim, mute me
14:03:40 <Zakim> TallTed should now be muted
14:03:59 <Arnaud> zakim, who's on the phone?
14:03:59 <Zakim> On the phone I see Sandro, Arnaud, Roger, +43.660.274.aaaa, SteveS, TallTed (muted), Ashok_Malhotra
14:04:04 <TallTed> TallTed has changed the topic to: LDP WG: http://www.w3.org/2012/ldp - next agenda: http://www.w3.org/2012/ldp/wiki/Meetings:Telecon2014.03.24
14:04:06 <Zakim> +JohnArwe
14:04:12 <sergio> Zakim,  +43.660.274.aaaa is me
14:04:12 <Zakim> +sergio; got it
14:04:41 <Zakim> +ericP
14:05:56 <Arnaud> zakim, who's on the phone?
14:05:56 <Zakim> On the phone I see Sandro, Arnaud, Roger, sergio, SteveS, TallTed (muted), Ashok_Malhotra, JohnArwe, ericP
14:07:41 <ericP> scribenick: ericP
14:08:00 <Arnaud> http://www.w3.org/2013/meeting/ldp/2014-03-17
14:08:00 <ericP> Proposal: Approve the minutes of the March 17 teleconf: http://www.w3.org/2013/meeting/ldp/2014-03-17
14:09:13 <SteveS> +1
14:09:28 <ericP> +1
14:09:43 <ericP> topic: next meeting
14:10:17 <SteveS> ericP did you remember ACTION-135 ?
14:10:27 <ericP> Arnaud: europe changes times so we'll back in synch between US and Europe
14:10:35 <ericP> topic: open actions
14:11:41 <bblfish> bblfish has joined #ldp
14:12:27 <Zakim> +bblfish
14:12:32 <ericP> topic: F2F5 Boston
14:13:05 <ericP> Arnaud: only a hand-full of folks have updated the wiki with their attendance expectations
14:13:18 <ericP> -> http://www.w3.org/2012/ldp/wiki/F2F5 F2F5 expected attendance
14:13:32 <ericP> topic: LDP spec
14:13:56 <ericP> Arnaud: sergio doubts that the link header can be used in a request
14:14:13 <bblfish> http://tools.ietf.org/html/rfc5988
14:14:19 <sandro> q+
14:14:31 <sergio> q+
14:14:44 <ericP> sergio: specified as a server header
14:14:56 <Arnaud> ack sandro
14:15:03 <Zakim> +??P16
14:15:09 <JohnArwe> q+ to mention the contents of 5988 section 5
14:15:13 <nmihindu> Zakim, ??P16 is me
14:15:13 <Zakim> +nmihindu; got it
14:15:19 <nmihindu> Zakim, mute me
14:15:19 <Zakim> nmihindu should now be muted
14:15:20 <bblfish> q+
14:15:29 <ericP> sandro: i asked mnot a week ago whether we could use link on a request
14:15:36 <ericP> mnot said sure, in a POST
14:15:38 <sandro> s/week/year
14:15:40 <Arnaud> ack sergio
14:15:40 <ericP> ... mnot said sure, in a POST
14:15:58 <JohnArwe> http://tools.ietf.org/html/rfc5988#section-5 specifically says: The Link entity-header field provides a means for serialising one or
14:15:58 <JohnArwe>    more links in HTTP headers.
14:16:02 <ericP> ... that's not a formal answer, but mnot's an authority
14:16:40 <Arnaud> ack JohnArwe
14:16:40 <Zakim> JohnArwe, you wanted to mention the contents of 5988 section 5
14:16:48 <ericP> sergio: it'd be nice to have an authoritative answer
14:16:52 <JohnArwe> section 4.1 says: However, they can specify the
14:16:52 <JohnArwe>    behaviours and properties of the target resource (e.g., allowable
14:16:52 <JohnArwe>    HTTP methods, request and response media types that must be
14:16:52 <JohnArwe>    supported).
14:17:52 <ericP> JohnArwe: 5.1 has the usage and abstraction of the link header. says that link is an "entity header"
14:18:02 <sandro> "The Link entity-header field provides a means" ....
14:18:25 <Arnaud> ack bblfish
14:18:40 <bblfish>  It is semantically equivalent to the   <LINK> element in HTML, as well as the atom:link feed-level element
14:18:40 <bblfish>    in Atom [RFC4287]
14:18:41 <ericP> ... "entity-header" is a superclass of req and resp headers (if you trust the editors to have been accurate in that phrasing)
14:19:19 <ericP> bblfish: you can add link to atom entries and you can post atom entries (but i suppose that doesn't say much)
14:19:20 <bblfish> Link: <> rel=type
14:19:32 <bblfish> <>
14:19:43 <ericP> ... in a post, you might want to specify that the subject of the link header is the document that will be created
14:20:18 <ericP> ... just as we said that <> in Turtle refers to the doc to be created
14:20:23 <bblfish> Link: <Container>, rel=type
14:20:59 <ericP> Arnaud: have we convinced ourselves that the link header is ok in POST reqs?
14:21:34 <ericP> bblfish: it should also be available on PUT
14:21:48 <ericP> JohnArwe: yes, to cover the create case
14:22:23 <JohnArwe> 5988 section 5.2 defines the (default) context URI in terms of "the requested resource", so POSTs would need to include an explicit context URI.  When create is the semantic, probably do need to define <> as the default context along with our interpretation of <>
14:23:12 <ericP> sergio: [satisfied]
14:23:18 <sergio> +1
14:23:39 <ericP> JohnArwe: the default URI is defined in terms of the requested resource
14:23:56 <bblfish> +1
14:24:02 <bblfish> who spoke?
14:24:06 <ericP> ... so on creates, we need a way to provide a URI that refers to the to-be-created resource, as we already did in the Turtle representaiton
14:24:11 <bblfish> +1 to that
14:24:18 <ericP> s/on creates/on create/
14:24:20 <Arnaud> q?
14:24:49 <nmihindu> JohnArwe, but our Link headers will be absolute URIs. aren't they ? so the context does not matter that.
14:26:24 <JohnArwe> nandana, on the create input representation, what context URI do you think corresponds to the "to be created" resource?  since the sender cannot know that until after receiving the create response, how does the creator supply a "valid" link header context uri?
14:26:54 <ericP> bblfish: on PUT/create can you create LDPCs or just LDPRs
14:27:25 <ericP> JohnArwe: we don't require creation of LDPC, but if the server is willing...
14:28:06 <ericP> Arnaud: if it's not a PUT/create, can you change the type of the resource?
14:28:18 <ericP> sandro: you can *ask* to change the type of the resource
14:29:12 <ericP> ... i take the phrase "requested resource" to be lazy writing, means the request-URI
14:29:35 <sandro> (or thing denoted by the request-URI)
14:29:44 <bblfish> cool
14:30:34 <JohnArwe> trackbot, johnarwe to look at required updates for 5988 + LInk header on creates
14:30:34 <trackbot> Sorry, JohnArwe, I don't understand 'trackbot, johnarwe to look at required updates for 5988 + LInk header on creates'. Please refer to <http://www.w3.org/2005/06/tracker/irc> for help.
14:30:39 <nmihindu> JohnArwe, I was rather think about rel='type' Link headers. For those, the context URI is irrelevant, isn't it. I thought it matters when we use relative URI in the Link header. But may be I am missing something.
14:30:50 <ericP> ACTION: JohnArwe to compose mail saying what we need to add to the spec for context URI in the link header on PUT and POST
14:30:50 <trackbot> Error finding 'JohnArwe'. You can review and register nicknames at <http://www.w3.org/2012/ldp/track/users>.
14:31:00 <ericP> ACTION: John to compose mail saying what we need to add to the spec for context URI in the link header on PUT and POST
14:31:00 <trackbot> Created ACTION-136 - Compose mail saying what we need to add to the spec for context uri in the link header on put and post [on John Arwe - due 2014-03-31].
14:32:44 <ericP> SteveS: we should reqply to sergio's request for clarity
14:32:53 <ericP> topic: Multiple Named Graphs
14:33:40 <ericP> Arnaud: betehess pressured us to add text that resources are bound by some graph. met resistance in public-ldp-comments
14:34:04 <ericP> sandro: i don't understand what the commentor wants.
14:34:26 <ericP> ... i don't think we need to respond before the commentor replies
14:34:50 <ericP> Arnaud: you seem unhappy with the named graphs text
14:35:08 <Zakim> -bblfish
14:35:14 <bblfish> damn
14:35:35 <ericP> bblfish: i work with Reto who wants to put everything in one graph
14:35:37 <Arnaud> Reto Gmür
14:35:59 <Zakim> +bblfish
14:36:01 <JohnArwe> I think he's just reading what he sees as conflicting things: (1) you must use named graphs ... I'm not convinced we do say that (2) you must support turtle ...which has no support for named graphs
14:36:14 <Arnaud> q?
14:36:51 <ericP> ... i spoke about a debate where folks wanted to put the membership triples in the data in the same graph for a direct container
14:37:21 <ericP> ... even if all the data is in one big graph behind the server, the client can only get views of that graph
14:37:34 <ericP> SteveS
14:37:56 <ericP> SteveS: can he write some SPARQL to retract those triples.
14:38:13 <ericP> ... if he wants to implement in one graph, that's up to him
14:38:44 <JohnArwe> I just searched editor's draft, I think he's just reading "corresponds to" as MUST
14:39:30 <ericP> topic: Paging spec
14:39:40 <sandro> issue-94?
14:39:40 <trackbot> issue-94 -- Stable vs lossy paging -- open
14:39:40 <trackbot> http://www.w3.org/2012/ldp/track/issues/94
14:39:58 <Arnaud> last week's proposal: add Sandro's requirement that client MUST get every triple in  intersect(g0, g1, ... gn) and MAY get any triples in the union (g0, g1,  .. gn), where STgx represents the graph at different times while the  client is traversing the pages
14:40:01 <sandro> (I'd call it "lossless" not "stable", myself.    :-)
14:40:20 <MiguelAraCo> MiguelAraCo has joined #ldp
14:40:50 <sandro> Gn is the state of the paged graph at the time the client does paging request number n.
14:41:06 <ericP> Arnaud: this keeps one from accidentally skipping triples that were there the whole time
14:41:27 <ericP> ... strawpoll largely supportive but some folks said they needed to test
14:41:38 <TallTed> it's neither lossless (in my mind, lossy includes DELETEs as well as INSERTs) nor stable...
14:41:58 <JohnArwe> @nandana: "I was rather think about rel='type' Link headers. For those, the context URI is irrelevant, isn't it"  Very Relevant. Using the RDF analogy, the context uri is the subject.  Saying that we're relying on the 5988 default being correct is very different from saying its value is irrelevant.
14:43:02 <ericP> ericP: so what happens when someone overfills a middle page in a container?
14:43:26 <ericP> sandro: cheap thing is to retire the URL for split pages, answering 410
14:44:22 <ericP> ... more helpful would be to @@2
14:44:29 <JohnArwe> +1 to ted's pt
14:45:05 <ericP> ... my theory is that between those two choices, we don't have serious performance vulnerabilities
14:45:58 <ericP> sandro: will anyone make tangible steps on this before CR?
14:46:10 <ericP> SteveS: maybe the 2nd week of April
14:47:28 <TallTed> I cannot urge strongly enough that folks read up on DBMS cursors and transaction isolation... cursors cover paging; isolation covers the "how do other client interactions with this server impact my own interaction?"
14:47:28 <TallTed> the "lowest level" of each of these is roughly what I think we're moving towards requiring, but we're using very different terminology to circle around to reinventing those DBMS features
14:47:33 <Ashok> q+
14:47:49 <Arnaud> ack Ashok
14:48:10 <sandro> PROPOSAL: Change LDP paging to be "lossless", defined as:  client who pages all the way through (forward or backward) MUST be sent every triple in  intersect(g0, g1, ... gn) and MAY get any triples in the union (g0, g1,  .. gn), where Gn is the paged graph at the time client makes request n
14:48:50 <TallTed> -1 "lossless"
14:48:58 <sandro> PROPOSAL: Change LDP paging to be "lossless", defined as:  client who pages all the way through (forward or backward) MUST be sent every triple in  intersect(g0, g1, ... gn) and MAY get any triples in the union (g0, g1,  .. gn), where Gi is the paged graph at the time client makes request i
14:49:20 <TallTed> q+
14:49:25 <TallTed> Zakim, unmute me
14:49:25 <Zakim> TallTed should no longer be muted
14:49:53 <ericP> Ashok: could you spell this out in a couple pages: what if the page is full, what if you're past the page, etc.
14:49:56 <Arnaud> ack TallTed
14:50:29 <ericP> TallTed: this is reinventing scrollable cursors and transaction isolation and blurring them in a bad way
14:50:36 <nmihindu> JohnArwe, thanks. I see the problem now and also why anchor parameter won't help because the URI of the created resource is not known. I wonder whether we can override the default for this case for the POST create in the LDP spec,
14:51:36 <sandro> PROPOSAL: Change LDP paging to be "lossless", defined as:  client who pages all the way through (forward or backward) MUST be sent every triple in  intersect(g0, g1, ... gn) and MAY get any triples in the union (g0, g1,  .. gn), where Gi is the paged graph at the time client makes request i.   Servers MAY refuse to serve any page because they'd want to implement lossless paging in certain cases; if they do so, they MUST use a 410 GONE.
14:51:45 <ericP> ... we should roughly re-word scrollable cursors and transaction isolation in our world
14:51:50 <sandro> q?
14:52:05 <ericP> ... by munging them together we set ourselves back a lot
14:52:26 <ericP> Ashok: would it help to have an email which spelled out how SQL handles this?
14:53:04 <ericP> Arnaud: sandro, please put your proposal in email. TallTed, please respond with your proposal.
14:53:49 <ericP> ... Ashok, once it's in writing, you can look at it with a SQL eye.
14:53:51 <sergio> q+
14:53:57 <ericP> topic: Test suite
14:54:02 <ericP> Arnaud
14:54:20 <SteveS> Note, we gathered information in the past http://www.w3.org/2012/ldp/wiki/Testing
14:54:21 <ericP> Arnaud: joined the call today [to talk about the test suite]
14:55:09 <Zakim> -bblfish
14:55:11 <nmihindu> https://dvcs.w3.org/hg/ldpwg/raw-file/b3683634c29f/Test%20Cases/LDP%20Test%20Cases.html
14:55:23 <ericP> ... we have a testing page, a high-level description by Raúl, test suite by ericP
14:55:41 <nmihindu> I would ping and ask him to reply in email
14:55:59 <ericP> ... nmihindu, can we get Raúl's attention again or should we re-assign?
14:56:11 <Arnaud> ack sergio
14:56:34 <ericP> sergio: it would be a cool to have a measure of where we are going
14:56:56 <ericP> ... something like what sandro did, something with SPARQL or whatever
14:57:19 <ericP> Arnaud: we have domain-specific servers which won't support vanilla tests
14:57:23 <JohnArwe> @sandro: the impln I was most worried about says they're already meeting the kind of guarantee you're proposing.  the impln currently would actually provide something a tad stronger, but that tad is not really part of the i/f so it's not a "guar" technically
14:57:34 <ericP> ... but we can still develop vanilla tests.
14:57:41 <Zakim> +bblfish
14:58:20 <ericP> ... and folks may adapt the vanilla tests to domain-specific server tests
14:58:33 <Arnaud> q?
14:58:49 <ericP> sergio: once we define the test cases, the tests will be pretty easy
14:58:54 <SteveS> q+
14:59:34 <Arnaud> ack Steves
14:59:45 <ericP> Arnaud: ericP, you have a test suite
14:59:47 <ericP> ericP
14:59:58 <ericP> ericP: mine was quite vanilla
15:00:37 <ericP> SteveS: should we have a task force of folks who will handle testing?
15:00:48 <sandro> +1 I want to be involved
15:00:52 <ericP> Arnaud: volunteers?
15:01:01 <sergio> +1, I can help
15:01:08 <SteveS> +1 I will be involved (or have someone indirectly working with me)
15:02:46 <bblfish> q+
15:03:06 <Arnaud> ack bblfish
15:03:46 <Zakim> -Roger
15:04:13 <SteveS> roger, nmihindu ... did you have any primer update?
15:05:52 <Zakim> -JohnArwe
15:06:23 <ericP> -> https://github.com/ericprud/SWObjects/blob/sparql11/tests/test_LDP.cpp ericP's (now ancient) generic LDP client/server tests
15:06:27 <SteveS> great, the primer is very valuable...looking forward to updates
15:06:28 <nmihindu> SteveS, we are working on the primer offline at the moment and we will do the updates on the repo soon.
15:06:54 <JohnArwe> sorry realized another mtg I need to be on had started
15:08:04 <sergio> -> https://git-wip-us.apache.org/repos/asf?p=marmotta.git;a=blob;f=platform/marmotta-ldp/src/test/java/org/apache/marmotta/platform/ldp/webservices/LdpWebServiceTest.java;h=f1109467ffae40cea95d92db926bee70f8650ae8;hb=ldp current marmotta ldp tests
15:08:14 <Zakim> -bblfish
15:08:17 <Zakim> -Ashok_Malhotra
15:08:18 <Zakim> -nmihindu
15:08:18 <Zakim> -Sandro
15:08:19 <Zakim> -sergio
15:08:19 <Zakim> -Arnaud
15:08:20 <Zakim> -TallTed
15:08:20 <Zakim> -SteveS
15:08:22 <Zakim> -ericP
15:08:23 <Zakim> SW_LDP()10:00AM has ended
15:08:23 <Zakim> Attendees were Sandro, Roger, Arnaud, SteveS, Ashok_Malhotra, TallTed, JohnArwe, sergio, ericP, bblfish, nmihindu
17:04:59 <Zakim> Zakim has left #ldp
17:36:37 <bblfish> bblfish has joined #ldp
17:45:18 <Arnaud1> Arnaud1 has joined #ldp
21:19:55 <Arnaud> Arnaud has joined #ldp
23:22:49 <SteveS> SteveS has joined #ldp