IRC log of rdf-wg on 2013-02-20

Timestamps are in UTC.

15:58:39 [RRSAgent]
RRSAgent has joined #rdf-wg
15:58:39 [RRSAgent]
logging to
15:58:41 [trackbot]
RRSAgent, make logs world
15:58:41 [Zakim]
Zakim has joined #rdf-wg
15:58:43 [trackbot]
Zakim, this will be 73394
15:58:43 [Zakim]
ok, trackbot; I see SW_RDFWG()11:00AM scheduled to start in 2 minutes
15:58:44 [trackbot]
Meeting: RDF Working Group Teleconference
15:58:44 [trackbot]
Date: 20 February 2013
15:59:53 [davidwood1]
Zakim, who is here?
15:59:53 [Zakim]
SW_RDFWG()11:00AM has not yet started, davidwood1
15:59:54 [Zakim]
On IRC I see RRSAgent, gkellogg, tbaker, Guus, TallTed, ivan, SteveH, AndyS, manu, davidwood1, yvesr, mischat, manu1, trackbot, sandro, ericP
16:00:21 [AndyS]
zakim, this is 73394
16:00:21 [Zakim]
ok, AndyS; that matches SW_RDFWG()11:00AM
16:00:22 [PatH]
PatH has joined #rdf-wg
16:00:36 [AndyS]
zakim, who is on the phone?
16:00:36 [Zakim]
On the phone I see [IPcaller], +1.540.898.aaaa, [IPcaller.a]
16:00:39 [ScottB]
ScottB has joined #rdf-wg
16:00:45 [AndyS]
zakim, IPCaller.a is me
16:00:46 [Zakim]
+AndyS; got it
16:00:48 [Zakim]
16:00:55 [gkellogg]
zakim, I am ??P10
16:00:56 [Zakim]
+gkellogg; got it
16:00:56 [Zakim]
16:01:00 [davidwood]
Zakim, aaaa is me
16:01:00 [Zakim]
+davidwood; got it
16:01:04 [TallTed]
Zakim, OpenLink_Software is temporarily me
16:01:04 [Zakim]
+TallTed; got it
16:01:12 [TallTed]
Zakim, mute me
16:01:13 [Zakim]
TallTed should now be muted
16:01:14 [Zakim]
+ +31.30.889.aabb
16:01:25 [Zakim]
+ +1.617.324.aacc
16:01:27 [Guus]
zakim, +31.30.889.aabb is me
16:01:27 [Zakim]
+Guus; got it
16:01:51 [Guus]
zakim, mute me
16:01:51 [Zakim]
Guus should now be muted
16:01:59 [Zakim]
- +1.617.324.aacc
16:02:06 [davidwood]
Zakim, who is here?
16:02:06 [Zakim]
On the phone I see tbaker, davidwood, AndyS, gkellogg, TallTed (muted), Guus (muted)
16:02:08 [Zakim]
On IRC I see ScottB, PatH, Zakim, RRSAgent, gkellogg, tbaker, Guus, TallTed, ivan, SteveH, AndyS, manu, davidwood, yvesr, mischat, manu1, trackbot, sandro, ericP
16:02:26 [markus]
markus has joined #rdf-wg
16:02:34 [Zakim]
+ +1.617.324.aadd
16:02:35 [Zakim]
16:02:41 [ivan]
zakim, aadd is ivan
16:02:41 [Zakim]
+ivan; got it
16:02:44 [markus]
zakim, what's the code
16:02:44 [Zakim]
I don't understand 'what's the code', markus
16:02:47 [manu]
zakim, I am ??P26
16:02:47 [Zakim]
+manu; got it
16:02:53 [markus]
zakim, code?
16:02:54 [Zakim]
the conference code is 73394 (tel:+1.617.761.6200, markus
16:03:22 [Zakim]
16:03:25 [cgreer]
cgreer has joined #rdf-wg
16:03:32 [davidwood]
Zakim, pick a victim
16:03:32 [Zakim]
Not knowing who is chairing or who scribed recently, I propose Guus (muted)
16:03:42 [davidwood]
Zakim, pick a victim
16:03:42 [Zakim]
Not knowing who is chairing or who scribed recently, I propose gkellogg
16:03:45 [Zakim]
16:03:53 [davidwood]
Zakim, pick a victim
16:03:53 [Zakim]
Not knowing who is chairing or who scribed recently, I propose tbaker
16:03:58 [markus]
zakim, ??P32 is me
16:03:58 [Zakim]
+markus; got it
16:04:00 [Zakim]
+ +1.408.992.aaee
16:04:08 [davidwood]
Zakim, pick a victim
16:04:08 [Zakim]
Not knowing who is chairing or who scribed recently, I propose gkellogg
16:04:13 [pfps]
pfps has joined #rdf-wg
16:04:19 [davidwood]
Zakim, who is here
16:04:19 [Zakim]
davidwood, you need to end that query with '?'
16:04:27 [Arnaud]
Arnaud has joined #rdf-wg
16:04:29 [davidwood]
Zakim, who is here?
16:04:29 [Zakim]
On the phone I see tbaker, davidwood, AndyS, gkellogg, TallTed (muted), Guus (muted), ivan, manu, Sandro, markus, +1.408.992.aaee
16:04:31 [Zakim]
On IRC I see Arnaud, pfps, cgreer, markus, ScottB, PatH, Zakim, RRSAgent, gkellogg, tbaker, Guus, TallTed, ivan, SteveH, AndyS, manu, davidwood, yvesr, mischat, manu1, trackbot,
16:04:31 [Zakim]
... sandro, ericP
16:04:35 [Zakim]
+ +1.707.874.aaff
16:04:46 [cgreer]
zakim, aaff is me
16:04:46 [Zakim]
+cgreer; got it
16:04:57 [Zakim]
16:05:04 [SteveH]
Zakim, ??P35 is me
16:05:04 [Zakim]
+SteveH; got it
16:05:07 [AndyS]
scribenick: AndyS
16:05:10 [Zakim]
16:05:15 [pfps]
Zakim, aaee is me
16:05:15 [Zakim]
+pfps; got it
16:05:46 [AndyS]
topic: admin
16:05:48 [Zakim]
16:05:54 [davidwood]
PROPOSED to accept the minutes of the 13 February telecon:
16:06:03 [Zakim]
16:06:08 [AndyS]
16:06:14 [Souri]
Souri has joined #rdf-wg
16:06:15 [pfps]
They're beautiful.
16:06:24 [davidwood]
RESOLVED to accept the minutes of the 13 February telecon:
16:06:33 [davidwood]
Review of action items
16:06:33 [davidwood]
16:06:33 [davidwood]
16:06:40 [zwu2]
zwu2 has joined #rdf-wg
16:06:57 [AndyS]
27 open actions
16:07:14 [davidwood]
Topic: Extension request
16:07:17 [Zakim]
16:07:34 [ScottB]
zakim, Tony is temporarily me
16:07:34 [Zakim]
+ScottB; got it
16:07:38 [Zakim]
+ +1.650.265.aagg
16:07:52 [zwu2]
zakim, +1.650.265.aagg is me
16:07:52 [Zakim]
+zwu2; got it
16:07:52 [AndyS]
davidwood: request informally approved, waiting for formal confirmation
16:07:55 [zwu2]
zakim, mute me
16:07:55 [Zakim]
zwu2 should now be muted
16:09:15 [AndyS]
davidwood: design time over, now complete docs during the extension
16:09:47 [AndyS]
... docs to get out soon -- semantics, concepts.
16:10:07 [AndyS]
topic: Turtle
16:10:10 [AndyS]
16:10:11 [Guus]
zakim, unmute me
16:10:11 [Zakim]
Guus should no longer be muted
16:10:25 [AndyS]
davidwood: congrats to all
16:10:38 [davidwood]
ack AndyS
16:10:39 [PatH]
One quick request. There is a CSS script that COncepts apparently used. Where can I get the original of that? Send email offline.
16:11:22 [AndyS]
andys: what's the impl feedback process?
16:11:58 [davidwood]
Implementors should inform public-rdf-comments
16:13:01 [AndyS]
davidwood: a quick blog to note process and EARL reports.
16:13:03 [Guus]
16:13:15 [davidwood]
ack Guus
16:13:40 [davidwood]
Zakim, who is here?
16:13:40 [Zakim]
On the phone I see tbaker, davidwood, AndyS, gkellogg, TallTed (muted), Guus, ivan, manu, Sandro, markus, pfps, cgreer, SteveH, Arnaud (muted), Souri, PatH, ScottB, zwu2 (muted)
16:13:44 [Zakim]
On IRC I see zwu2, Souri, Arnaud, pfps, cgreer, markus, ScottB, PatH, Zakim, RRSAgent, gkellogg, tbaker, Guus, TallTed, ivan, SteveH, AndyS, manu, davidwood, yvesr, mischat, manu1,
16:13:44 [Zakim]
... trackbot, sandro, ericP
16:14:38 [AndyS]
gkellog: is it Eric's tests as well?
16:14:47 [davidwood]
ACTION: davidwood to write a blog post announcing Turtle to CR
16:14:47 [trackbot]
Created ACTION-230 - Write a blog post announcing Turtle to CR [on David Wood - due 2013-02-27].
16:14:58 [gkellogg]
16:15:48 [davidwood]
ACTION: davidwood to create a wiki page to track Turtle CR comments and notify the editors.
16:15:48 [trackbot]
Created ACTION-231 - Create a wiki page to track Turtle CR comments and notify the editors. [on David Wood - due 2013-02-27].
16:15:58 [AndyS]
gkellogg: please rerun tests to be upto date
16:16:09 [AndyS]
davidwood: need to track CR comments
16:16:26 [TallTed]
Zakim, unmute me
16:16:26 [Zakim]
TallTed should no longer be muted
16:17:11 [AndyS]
ivan: all changes now need to be tracked
16:17:20 [AndyS]
... we're in process mode now
16:17:49 [AndyS]
topic: semantics
16:18:18 [davidwood]
▪Should we allow blank nodes to be used as graph names? That is, allow a graph name to be both (IRI, graph), and (blank node, graph).
16:18:18 [davidwood]
NB: JSON-LD has a preference to allow blank nodes as graph names.
16:18:18 [davidwood]
From Manu: PROPOSAL: Allow blank nodes to be used as graph names. Specifically, allow associating (IRI, graph) and (blank node, graph) when naming graphs.
16:19:16 [AndyS]
manu: review of JSON-LD status. Has bnode ids fro graphs and also unlabelled graphs.
16:19:24 [TallTed]
Zakim, mute me
16:19:24 [Zakim]
TallTed should now be muted
16:19:27 [AndyS]
... long discussions
16:19:52 [AndyS]
... part 2 -- denotation of graph (values)
16:20:17 [AndyS]
... only interpretation possible
16:20:29 [AndyS]
(Andy notes that claim is false)
16:20:47 [pfps]
The method used for containers works for me. I'm not in favour of making any further changes to the semantics.
16:21:24 [AndyS]
PatH: opportunity to add the semantic condition that bNode denotes the graph.
16:21:53 [manu]
q+ to state that Web Payments is going to assert that IRIs and blank node identifiers denote graphs. We have to.
16:22:20 [pfps]
16:22:20 [AndyS]
... bnodes for labels can be used to put metadata into datasets
16:23:01 [AndyS]
???: skolemization would break that condition?
16:23:21 [AndyS]
path: not really - we don't stop IRIs denoting graphs
16:23:56 [AndyS]
16:25:07 [AndyS]
sandro: neutral on proposal
16:25:39 [manu]
I think that somebody using an IRI to name a graph previously, where the IRI doesn't denote the graph, did something non-standard and we don't need to support that in RDF Concepts 1.1 (we are breaking some deployments, but for the greater good of the Web)
16:25:49 [SteveH]
16:26:03 [davidwood]
ack manu
16:26:03 [Zakim]
manu, you wanted to state that Web Payments is going to assert that IRIs and blank node identifiers denote graphs. We have to.
16:26:08 [pfps]
even if you are using the skolem namespace I don't think that you can infer the denotation
16:26:54 [AndyS]
manu: web payments will use denotation
16:27:02 [ivan]
16:27:12 [davidwood]
ack pfps
16:27:16 [PatH]
pfps, you cany infer it but we can impose it as a condition.
16:27:29 [manu]
+100 to pfps!
16:27:34 [PatH]
16:27:36 [manu]
-1 for it being too late :)
16:27:43 [AndyS]
pfps: not done properly; it's too late.
16:27:46 [Zakim]
16:28:49 [manu]
q+ to ask pfps if he'd be okay with blank nodes as graph names, but -1 on denotation?
16:29:00 [zwu2]
+1 to pfps
16:29:05 [SteveH]
16:29:11 [ivan]
16:29:47 [manu]
It's a big problem for us!
16:29:55 [manu]
(for the web payments work)
16:30:01 [manu]
it's a big problem for digital signatures.
16:30:10 [manu]
it's a big problem for RDF Dataset Normalization.
16:30:18 [AndyS]
manu - do you recognize that forcing one decision is a problem for others?
16:30:36 [manu]
Which decision am I forcing, andys?
16:30:38 [SteveH]
manu, we do all that stuff, and we don't use blank graphs
16:30:52 [manu]
SteveH, then what's the solution?
16:31:01 [SteveH]
16:31:02 [Souri]
+1 against use of bNode for graphs
16:31:14 [davidwood]
ack SteveH
16:31:57 [pfps]
If we put graph name denotation into RDF then the semantics has to be expanded to include graphs as a new datatype
16:32:04 [AndyS]
manu - you asked for denotation only in JSON-LD. Other people do different things. You case is fine - it's one amongst several - other uses are important to their users.
16:32:18 [AndyS]
SteveH: taken out of 3Store
16:32:22 [PatH]
its not a change to how rdf works.
16:32:48 [davidwood]
ack ivan
16:33:18 [AndyS]
ivan: time issues
16:34:56 [davidwood]
ack manu
16:34:56 [Zakim]
manu, you wanted to ask pfps if he'd be okay with blank nodes as graph names, but -1 on denotation?
16:35:46 [AndyS]
manu: 2 proposals - one syntax, one semantics
16:36:05 [AndyS]
.. can accept first, not the second
16:36:11 [pfps]
If SPARQL doesn't allow blank nodes as names in datasets, then I don't see a need for us to.
16:36:23 [PatH]
then i disagree. bnodes the dont denote really arre meaningless.
16:36:24 [AndyS]
ivan: denotation is controversal
16:36:43 [AndyS]
manu: new evidence
16:36:54 [PatH]
peter, sparql does allow it, i gather.
16:37:15 [PatH]
16:37:38 [ivan]
zakim, who is noisy?
16:37:46 [AndyS]
SPARQL does not but it's no big deal. One line change. (someone set things up for future posibilities :-)
16:37:50 [Zakim]
ivan, listening for 10 seconds I heard sound from the following: AndyS (39%), manu (99%)
16:38:25 [SteveH]
manu, several people have suggested perfectly workable solutions
16:38:44 [pfps]
the method proposed that gets rid of blank nodes doesn't require minting new IRIs, you just have an infinite number of them pre-allocated (like rdf:_<n>) and choose in order
16:39:01 [AndyS]
ivan: propose to close second proposal
16:39:27 [AndyS]
q+ to say what SPARQL actually says
16:39:38 [manu]
steveh, no, they haven't - we've spent a great deal of time trying to apply those solutions - they are half-baked, or don't work.
16:39:51 [manu]
pfps: Isn't that a new IRI scheme?
16:40:17 [gkellogg]
andys: a definition of a dataset in SPARQL says it's an IRI; the mechanics, translation to algebra and evaluation are neutral, because you can't write it.
16:40:19 [manu]
pfps: graph:_nnn <-- new IRI scheme, most of the folks on this mailing list didn't want that, right?
16:40:30 [gkellogg]
… Blank nodes would be variables.
16:40:54 [gkellogg]
… SPARQL would work with BNodes for properties, it doesn't interpret them.
16:41:03 [SteveH]
but there are no (real) blank node in sparql
16:41:31 [gkellogg]
… I don't know if actual engines would blow up or not, but structurally, it wouldn't make a difference.
16:41:40 [AndyS]
16:41:40 [SteveH]
4store and 5store would blow up, that section of the index (bnode as graph ID) doesn't exist
16:41:51 [davidwood]
ack PatH
16:41:57 [AndyS]
16:42:35 [SteveH]
16:43:00 [AndyS]
davidwood: we have the freedom but not the time to deviate from SPARQL
16:46:03 [ivan]
16:46:10 [cgreer]
cgreer has joined #rdf-wg
16:46:33 [AndyS]
manu: can JSON-LD use bNode id for graphs in datasets?
16:47:05 [davidwood]
ack ivan
16:47:18 [SteveH]
IIUC manu asked about nquads… that's not OK
16:47:21 [SteveH]
for me
16:47:46 [AndyS]
ivan: maybe a way out but may have consequences outside JSON-LD.
16:48:04 [gkellogg]
BNodes for property IDs in JSON-LD will already make these other applications blow up.
16:48:05 [AndyS]
... no reason for deciding one over the other.
16:49:54 [AndyS]
davidwood: try to find a loose framework that allows variation.
16:51:08 [SteveH]
that must already happen, because of bNode predicates
16:51:23 [SteveH]
that's not legal in RDF and can't be serialised
16:51:23 [AndyS]
ivan: suggest a warning in JSON-LD about the consequences outside JSON-LD
16:51:27 [markus]
the warnings are already there: and again here
16:52:01 [AndyS]
davidwood: how bad is this?
16:53:16 [AndyS]
markus - quite. Various tricky points there - e.g. lists in data model (wish that were true) We just don't ask!
16:53:36 [PatH]
16:53:41 [AndyS]
manu: causes an issue - need to label the graph in some way
16:53:46 [markus]
andys, I know :-)
16:54:08 [AndyS]
... reinventing bnode ids
16:54:41 [PatH]
Let me withdraw my blocking vote decision here. I dont like semantics=free bnodeIDs, but I think compatibility with JSON is more important than my aesthetics. SO I will vote for the syntax without the sematnics. <grinds teeth.>
16:55:36 [AndyS]
ivan: will break other use cases
16:58:11 [PatH]
The key point for Steve seems to be that he did have this but was ASKED to remove it. WHo asked him and what were their reasons?
16:59:49 [PatH]
Ivan, there are recent emails showing that SPARQL in fact works fine with this case.
17:00:06 [PatH]
+1 to Manu
17:00:09 [SteveH]
PatH, I don't see how that can be the case
17:00:30 [Souri]
I like the (alternate) proposal of minting a new IRI (it avoids the new requirement of allowing bnode identifying a graph)
17:00:32 [PatH]
Its not a change to RDF *at all*. It might be a change to SPARQL.
17:01:22 [SteveH]
PatH, what about the case where the bNode identified by _:aaa is used as a graph label, but doesn't appear in any graph?
17:02:04 [AndyS]
manu; new info - normalization, JSON-LD, documents without base URI
17:02:11 [AndyS]
manu: new info - normalization, JSON-LD, documents without base URI
17:02:32 [davidwood]
ack PatH
17:03:14 [PatH]
Steve, what about it? Its kind of silly (semantically) but harmless.
17:03:21 [manu]
New information to blank nodes as graph name labels - Did the WG consider RDF Dataset Normalization when you discussed this?
17:03:35 [Arnaud]
I have to drop off for another call, I find unfortunate that the number of incompatibilities between JSON-LD and RDF keeps increasing
17:03:38 [SteveH]
PatH, but it wouldn't be a node (and it's what manu is proposing to do, FWIW)
17:04:11 [manu]
Did the WG consider how the decision would affect JSON-LD developers, specifically how forcing them to use an identifier where they don't have to use one (when expressing a blank node)
17:04:16 [PatH]
religious = doctrinal :-)
17:04:24 [AndyS]
ivan: don't know it means to have no base URI.
17:04:57 [Arnaud]
which isn't to say that we should let JSON-LD influence RDF, I find this slightly odd because I don't think the syntax should necessarily influence the model
17:05:07 [AndyS]
... normalization is significant but was worked on it elsewhere.
17:05:13 [Zakim]
17:05:32 [gkellogg]
I don't think that normalization actually requires that graph names include BNode identifiers, it just does if JSON-LD allows them, and that output should be normalized.
17:05:34 [AndyS]
... this WG has not picked up that work.
17:06:02 [AndyS]
... at this time and this point in process, it's hard to take that on.
17:06:02 [PatH]
SteveH, ?? of course it would be a node. Bnodes are nodes.
17:06:18 [SteveH]
PatH, well, only if they appear in a graph, surely?!
17:06:49 [AndyS]
manu: we have a solution. Need a direction from this WG
17:06:52 [davidwood]
17:06:59 [PatH]
IWhy? We are allowing IRIs to be labels. why no other kinds of node?
17:07:04 [AndyS]
.. bnode was the thing we choose at the time.
17:08:09 [SteveH]
it doesn't require a new URI scheme, you could use the skolem one
17:08:11 [pfps]
I don't see how this group is on the hook for doing anything here.
17:08:18 [AndyS]
manu: new URI scheme is an alternative but not as attractive.
17:08:44 [pfps]
17:09:35 [davidwood]
ack pfps
17:10:47 [SteveH]
IIRC, you're required to have a base URI, as per RFC 2396
17:10:53 [PatH]
The hostility isn't necessary.
17:13:26 [PatH]
+1 to manu
17:13:39 [AndyS]
davidwood: if docs standardized as currently stated - manu - what happens?
17:14:23 [AndyS]
manu: we will use specific IRIs for the dataset.
17:14:23 [PatH]
Thats skolemizing, in fact.
17:16:22 [AndyS]
davidwood: will skolemization work for you?
17:16:51 [AndyS]
manu: will reply on the list
17:17:04 [PatH]
Well not quite.
17:17:39 [TallTed]
TallTed has joined #rdf-wg
17:18:07 [manu]
AndyS: You can't do that without taking inter-graph connectivity into account.
17:18:07 [PatH]
For the record, now we ahve extra time, how uch extra time do we have (The WG, not the call)?
17:18:15 [manu]
you can't just "hash a graph" to get a name.
17:18:22 [AndyS]
davidwood: wish we had more time ...
17:18:25 [Zakim]
17:18:26 [Zakim]
17:18:27 [Zakim]
17:18:28 [Zakim]
17:18:30 [Zakim]
17:18:31 [Zakim]
17:18:33 [Zakim]
17:18:34 [Zakim]
17:18:35 [Zakim]
17:18:36 [Zakim]
17:18:36 [Zakim]
17:18:38 [Zakim]
17:18:40 [Zakim]
17:18:43 [Zakim]
17:18:57 [Zakim]
17:19:11 [AndyS]
manu -if you want to denote the value, you can. In saying that I suspect you are not denoting the graph but an instance/usage of a graph
17:19:26 [SteveH]
manu, my issue is that a lot of this is specific to your implementation - you could have made other choices and ended up with different requests to the group - I see that you didn't, but you're not appreciating the (real) cost to other organisations to support your impl. choices
17:19:48 [AndyS]
17:19:59 [AndyS]
The real cost is education.
17:20:04 [manu]
AndyS: we are not trying to denote the value of the graph, though
17:20:21 [SteveH]
and complication, and plain off-ness
17:20:24 [SteveH]
17:20:33 [manu]
SteveH: "our implementation" is the RDF Dataset Normalization algorithm and JSON-LD (two specifications, not implementations of systems)
17:20:51 [AndyS]
manu - you asked the WG that the bnode does denote the value. This is the source of confusion.
17:21:25 [manu]
SteveH: I fully understand that there are other implementations out there that do things a different way, and that's fine - I'm trying to figure out how the interoperability picture works.
17:21:29 [SteveH]
manu, that's only partially true, but sure
17:21:43 [manu]
SteveH: If I didn't care about that, we would've just gone ahead and done this w/o consulting the group.
17:22:05 [manu]
AndyS: No, I didn't talk about denoting at all - PatH brought that up, and I thought it was a fair point and asked that it be put on the agenda.
17:22:10 [SteveH]
demanding breaking changes isn't consulting in my book
17:22:27 [manu]
SteveH: When did I demand anything?
17:22:37 [SteveH]
in the last 75 mins
17:22:56 [manu]
SteveH: Well, if you got that impression, that's certainly not what I was trying to express.
17:23:03 [AndyS]
you asked for the label to name the graph. A graph is a value. Two graphs are the same if they are the same set of triples. You seem to want something different which is fine but what you asked for bnodes naming graph as values.
17:23:46 [SteveH]
manu, you said a lot of things along the lines of, we require this feature
17:23:53 [SteveH]
it came across as a demand
17:23:57 [SteveH]
to me at least
17:24:20 [manu]
When did I say we 'require' the feature?
17:24:41 [manu]
I went to a great deal of trouble to try and outline the options that would/wouldn't work for us - and organize them in least bad to most bad (from our perspective).
17:24:43 [AndyS]
You also asked for IRIs to denote graphs ... no room for other use cases
17:26:05 [manu]
AndyS: re: IRIs to denote graphs - It's my opinion that doing anything else creates a great deal of confusion, and potentially disastrous repercussions in the financial use case we're working on. I'm entitled to that opinion, and entitled to voice it in this group, aren't I?
17:26:42 [SteveH]
yes, but there's plenty of counter-evidence
17:27:05 [SteveH]
you are not the only company working in the finance space, and no-one else has an issue with the non semantics of datasets
17:27:15 [manu]
There is evidence that people use graph names in ways that are very dangerous, I don't think we should be blessing that behavior.
17:27:25 [SteveH]
the logical conclusion (reductio ad absurdum, sorry) is that you can't build finance systems with SQL
17:27:50 [AndyS]
What evidence? Pointers?
17:27:55 [manu]
SteveH: I'm not saying it's impossible - I'm saying it throws a wrench into the works that not all developers are going to be savvy about.
17:28:24 [SteveH]
but hardly any developers have encountered graph systems, why would it throw a wrench into anything?
17:28:43 [AndyS]
There is a lot of evidence in deployed, real, for-money systems for a different approach.
17:28:46 [manu]
AndyS: You gave one of the examples yourself - they construct a URI for the time a graph was retrieved and use that as the name of the graph.
17:29:57 [SteveH]
how is that dangerous?
17:29:59 [manu]
The better solution is to retrieve the graph, name it as a blank node, and then annotate the graph with the retrieval time, document hash, etc.
17:30:12 [SteveH]
I can assure you, that's not better
17:30:19 [SteveH]
not if you have any longevity
17:30:22 [manu]
It's dangerous because you have a bunch of IRIs in the system that don't dereference to anything, but use IRIs.
17:30:37 [SteveH]
that's common, and totally harmless
17:31:32 [manu]
Well, seems like we have different definitions of harmless.
17:31:51 [AndyS]
The key is the "retrieve" - name the retrieveal.
17:32:18 [SteveH]
I suspect the definitions are the same, but experiences differ
17:32:19 [AndyS]
(with a bnode if you want)
17:32:43 [manu]
name the graph with a bnode?
17:33:17 [manu]
Clearly, we're not moving forward with blank node identifiers for graphs names in this group... so what's the next best option in both of your opinion?
17:34:16 [SteveH]
well, you're the one that came up with the requirement
17:34:55 [SteveH]
I absolutely don't understand your reasoning - none of my experience of working in this field for many years led me to the same conclusions
17:35:01 [Zakim]
disconnecting the lone participant, ivan, in SW_RDFWG()11:00AM
17:35:02 [Zakim]
SW_RDFWG()11:00AM has ended
17:35:02 [Zakim]
Attendees were +1.540.898.aaaa, AndyS, tbaker, gkellogg, davidwood, TallTed, +1.617.324.aacc, Guus, +1.617.324.aadd, ivan, manu, Sandro, markus, +1.408.992.aaee, +1.707.874.aaff,
17:35:02 [Zakim]
... cgreer, SteveH, Arnaud, pfps, Souri, PatH, ScottB, zwu2
17:35:05 [SteveH]
so I don't think I can offer an sensible advice
17:35:18 [SteveH]
also, I have to go :)
17:35:20 [SteveH]
bye all
17:35:20 [manu]
If it were you, how would you name the graphs?
17:35:48 [AndyS]
Recognize that other group have other requirements and make a proposal that is for everyone
17:36:43 [AndyS]
I don't think we can make progress when the proposals on the table exclude groups (because it's so late)
17:37:20 [manu]
I have 3 proposals, which one do you like the best: blank node identifiers for graph names, new IRI scheme for document-local identifiers, pick some magic IRI scheme that can be re-written when normalizing.
17:37:56 [AndyS]
which one do you think works for other use cases? What's the consensus suggestion?
17:39:11 [manu]
I thought all three of those worked, so did Pat, Gregg, and Markus - now you're asserting that it doesn't... so, the last two haven't been taken out - what do you think of those two solutions?
17:39:37 [manu]
AndyS: There is no consensus suggestion yet - that's what I'm trying to tease out of the RDF WG.
17:43:00 [markus]
manu, I think the easiest way forward is to mint payswarm-specific IRIs and assign them the meaning you need in the payswarm spec
17:43:49 [markus]
bnode IDs as graph names would be by far the best technical solution IMO but as we've seen today others disagree
17:44:17 [AndyS]
I'm waiting for a consensus creating proposal that speaks to all interests. When people suggest compromise, it is rejected (if it is acknowledged at all).
17:45:12 [manu]
How is "let's mint a new IRI scheme?" not seeking compromise?
17:45:25 [markus]
well... it's a yes-no decision whether bnodes are allowed as graph names or not...
17:45:40 [manu]
it's completely in line w/ RDF Concepts 1.1?
17:46:25 [markus]
manu, that's exactly the same as minting payswarm specific IRIs.. everything would work.. until you put it in a quad store and the IRIs collide
17:46:38 [markus]
so you have to find some way to generate IRIs that don't collide
17:47:12 [markus]
bnode are the only thing that, by definition, won't collide
17:47:13 [manu]
markus: Yes, well - that's the problem isn't it - all the IRIs that are being generated are being generated in a decentralized way.
17:47:24 [markus]
yes, it is
17:48:09 [markus]
alternative, you define the normalization algorithm in a way that works on top of JSON-LD instead of N-Quads
17:48:32 [manu]
and they all have to generate the IRI in the exact same way... we can just use the same algorithm we're using to name bnode-graph-labels now... but then, you put them in a quadstore and the quadstore thinks they're the same graph.
17:48:33 [AndyS]
All ecommerce replies on the same (the single use session keys)!
17:49:46 [manu]
AndyS: The problem with generating IRIs is that if the names that are generated on both ends aren't exactly the same, the digital signatures don't match.
17:49:59 [manu]
AndyS: that is - HTTP-like IRIs.
17:50:31 [manu]
AndyS: We can create a new class of IRI that is a document-local IRI, but then quadstores need to know that the IRI is special and to treat it like a document-local identifier when it is placed in a quad store.
17:51:03 [AndyS]
Going round in circles. Graphs are values
17:51:27 [manu]
That doesn't help me.
17:51:40 [AndyS]
Why not keep the choice of name inside the normalization algorithm?
17:51:58 [manu]
it already is.
17:52:13 [manu]
(for graphs that are not explicitly named in JSON-LD)
17:52:39 [manu]
The question is: What does the normalization algorithm name such a graph?
17:53:23 [manu]
Another way to cut it is to have JSON-LD assign a name to a graph before it goes to the RDF Dataset Normalization algorithm, and the RDF Dataset Normalization algorithm doesn't deal with naming at all (which is probably the right way to do it)
17:53:36 [manu]
but that still doesn't tell us what we should name the graph.
17:53:50 [manu]
Right now, we would probably go with something like <graph:1>
17:54:05 [markus]
manu, where get the unnamed named graphs get created in payswarm?
17:54:18 [markus]
I've just found and there are no named graphs
17:54:40 [manu]
It's not in this version of the spec at all, and those specs are very much out of date.
17:55:09 [manu]
right now, we have to deal with graph naming by puttng the signature in the default graph and assuming that the signature applies to the default graph.
17:55:36 [manu]
This will fall apart completely when RDFa Next adds graph support... well, unless RDFa Next requires that all graphs be named with an IRI.
17:55:39 [AndyS]
Again your use of graph based on structre in dataset is suggesting to me that you are just using the "gsnap" (which is what Pat seeks) -- so I think even if there is agreement now it will unravel as details get thrashed out.
17:56:32 [manu]
How is it going to unravel?
17:57:01 [AndyS]
RDFa needs graph labels to be locations (and maybe other things) else it will v hard to express anything without levels of indirection
17:57:41 [markus]
ok.. gotta go
17:57:44 [markus]
17:57:45 [AndyS]
because what pat supports and what you need for the alg aren't obviously the same - indeed I think they are actually different.
17:57:46 [manu]
What that translates into is asserting that people will need to do this: graph="#mygraph" instead of graph="" (bnode graph label)
17:58:58 [manu]
However, developers can already do this typeof="" (create a new blanknode subject) and don't have to pair ti with this about="#foo"
17:59:46 [AndyS]
ivan?ericP sandro? Problem with the minutes. Getting a 403
17:59:53 [manu]
As I said before, it's not a show stopper - it's just very strange that you can create blank nodes and attach properties to them with ease in RDF, but you can't create blank graphs and associate them with triples.
17:59:56 [AndyS]
18:13:13 [sandro]
AndyS, sorry about that bug. I did the temporary work-around (change "cgi" to "www" in that URL).
18:13:35 [sandro]
(they're phasing out the domain name, and I need to fix commonscribe to match.)
18:13:54 [AndyS]
NP - I looked for the latest change wiki page and hit buttons at random. Bit like spec writing.
18:16:28 [sandro]
:-) :-)
19:27:31 [Zakim]
Zakim has left #rdf-wg