IRC log of rdf-wg on 2011-04-20

Timestamps are in UTC.

14:25:49 [RRSAgent]
RRSAgent has joined #rdf-wg
14:25:49 [RRSAgent]
logging to
14:25:51 [trackbot]
RRSAgent, make logs world
14:25:51 [Zakim]
Zakim has joined #rdf-wg
14:25:53 [trackbot]
Zakim, this will be 73394
14:25:53 [Zakim]
ok, trackbot; I see SW_RDFWG()11:00AM scheduled to start in 35 minutes
14:25:54 [trackbot]
Meeting: RDF Working Group Teleconference
14:25:54 [trackbot]
Date: 20 April 2011
14:40:33 [AZ]
AZ has joined #rdf-wg
14:46:09 [SteveH]
SteveH has joined #rdf-wg
14:48:12 [OlivierCorby]
OlivierCorby has joined #rdf-wg
14:53:50 [ivan]
ivan has joined #rdf-wg
14:54:28 [cygri]
cygri has joined #rdf-wg
14:54:58 [Scott]
Scott has joined #rdf-wg
14:56:44 [mbrunati]
mbrunati has joined #rdf-wg
14:57:36 [Zakim]
SW_RDFWG()11:00AM has now started
14:57:44 [Zakim]
14:57:47 [SteveH]
Zakim, what's the code
14:57:47 [Zakim]
I don't understand 'what's the code', SteveH
14:57:51 [davidwood]
Chair: David Wood
14:57:57 [SteveH]
Zakim, what is the code?
14:57:57 [Zakim]
the conference code is 73394 (tel:+1.617.761.6200 tel:+ tel:+44.203.318.0479), SteveH
14:58:00 [davidwood]
SteveH: 73394
14:58:01 [mischat_]
mischat_ has joined #rdf-wg
14:58:08 [AndyS1]
AndyS1 has joined #rdf-wg
14:58:15 [gavinc]
gavinc has joined #rdf-wg
14:58:28 [Zakim]
14:58:38 [Zakim]
14:58:38 [AZ]
AZ has joined #rdf-wg
14:58:48 [Zakim]
14:58:58 [SteveH]
Zakim, ??P8 is me
14:58:58 [Zakim]
+SteveH; got it
14:59:05 [Zakim]
14:59:18 [mischat_]
zakim, ??P7 is me
14:59:18 [Zakim]
+mischat_; got it
14:59:20 [Scott]
Zakim Tony is me
14:59:22 [mischat_]
hello all
14:59:25 [gavinc]
zakim, code?
14:59:25 [Zakim]
the conference code is 73394 (tel:+1.617.761.6200 tel:+ tel:+44.203.318.0479), gavinc
14:59:51 [Zakim]
14:59:57 [Zakim]
14:59:58 [Zakim]
15:00:09 [davidwood]
Zakim, who is here?
15:00:09 [Zakim]
On the phone I see davidwood, SteveH, Tony, Sandro, mischat_, AxelPolleres, [IPcaller.a], [IPcaller]
15:00:11 [Zakim]
15:00:13 [AndyS]
zakim, IPcaller.a is me
15:00:13 [Zakim]
+AndyS; got it
15:00:14 [Zakim]
15:00:17 [mbrunati]
zakim, IPCaller is me
15:00:17 [Zakim]
+mbrunati; got it
15:00:20 [AndyS]
15:00:33 [Scott]
Zakim Tony is Scott
15:00:49 [Zakim]
15:00:51 [Zakim]
15:01:03 [Scott]
zakim, Tony is Scott
15:01:03 [Zakim]
+Scott; got it
15:01:11 [AZ]
zakim, Luca may be me
15:01:11 [Zakim]
+AZ?; got it
15:01:19 [AlexHall]
AlexHall has joined #rdf-wg
15:01:39 [cygri]
zakim, what's the coe?
15:01:39 [Zakim]
I don't understand your question, cygri.
15:01:44 [cygri]
zakim, what's the code?
15:01:44 [Zakim]
the conference code is 73394 (tel:+1.617.761.6200 tel:+ tel:+44.203.318.0479), cygri
15:01:59 [Zakim]
15:02:07 [ivan]
zakim, dial ivan-voip
15:02:07 [Zakim]
ok, ivan; the call is being made
15:02:09 [Zakim]
15:02:09 [Zakim]
15:02:23 [FabGandon]
FabGandon has joined #rdf-wg
15:02:24 [cygri]
zakim, mhausenblas is temporarily me
15:02:24 [Zakim]
+cygri; got it
15:02:29 [Zakim]
+ +34.67.92.aaaa
15:02:36 [NickH]
appologies, I can't make this meeting
15:02:47 [Zakim]
15:02:56 [webr3]
Zakim, i am ??P21
15:02:57 [Zakim]
+webr3; got it
15:03:11 [JFB]
zakim, i am +34.67.92.aaaa
15:03:13 [Zakim]
+JFB; got it
15:03:40 [zwu2]
zwu2 has joined #rdf-wg
15:03:55 [JFB]
zakim, mute me
15:03:55 [Zakim]
JFB should now be muted
15:03:58 [Zakim]
15:04:12 [cygri]
15:04:16 [PatH]
PatH has joined #rdf-wg
15:04:16 [Zakim]
15:04:23 [zwu2]
zakim, mute me
15:04:23 [Zakim]
zwu2 should now be muted
15:04:33 [davidwood]
Scribe: Olivier Corby
15:04:45 [davidwood]
ScribeNick: OlivierCorby
15:04:49 [Zakim]
+ +20598aabb
15:04:57 [danbri]
zakim, aabb is danbri
15:04:57 [Zakim]
+danbri; got it
15:05:09 [cygri]
AxelPolleres LOL
15:05:15 [Zakim]
15:05:24 [Souri]
Souri has joined #rdf-wg
15:05:50 [Zakim]
15:06:00 [davidwood]
PROPOSED to accept the minutes of the FTF1:
15:06:00 [davidwood]
15:06:00 [davidwood]
15:06:21 [OlivierCorby]
there is an issue
15:06:31 [mischat]
i haven't cleaned up my scribed session, I will definitely get round to that soon
15:06:34 [OlivierCorby]
note in the minutes
15:06:54 [Zakim]
15:06:56 [OlivierCorby]
two people would vote -1
15:07:01 [LeeF]
That doesn't seem like an issue with the minutes
15:07:26 [danbri]
(can the dissent be linked from the issue tracker as a hub?)
15:07:41 [FabGandon]
zakim, [Sophia] is me
15:07:41 [Zakim]
+FabGandon; got it
15:07:46 [LeeF]
Thanks very much to the scribes from F2F -- minutes are very helpful
15:07:52 [webr3]
one of the -1's was mine I know that
15:08:27 [tomayac]
LeeF: +1
15:09:00 [AZ]
this email is part of the thread:
15:09:09 [webr3]
15:09:25 [gavinc]
15:09:29 [AZ]
yes I did vote -1
15:09:33 [cygri]
15:09:55 [Zakim]
15:10:01 [Zakim]
15:10:01 [OlivierCorby]
put the URLs in the minutes
15:10:19 [MacTed]
Zakim, OpenLink_Software is temporarily me
15:10:19 [Zakim]
+MacTed; got it
15:10:22 [MacTed]
Zakim, mute me
15:10:22 [Zakim]
MacTed should now be muted
15:11:15 [OlivierCorby]
actions under review
15:11:26 [cygri]
davidwood, AZ:
15:11:35 [OlivierCorby]
action 20
15:11:35 [trackbot]
Sorry, bad ACTION syntax
15:11:40 [davidwood]
15:12:21 [mischat]
15:12:25 [mischat]
it looks ok to me ^^
15:12:43 [PatH]
we can close the opening.
15:12:59 [tomayac]
link correction:
15:13:10 [mischat]
sorry tomayac
15:13:22 [tomayac]
yours was webr3's great work
15:14:00 [cygri]
davidwood, and here's the -1 from webr3:
15:14:13 [OlivierCorby]
ACTION-27: Make sure the resolution to issue-12 gets into semantics document
15:14:14 [trackbot]
ACTION-27 Make sure the resolution to issue-12 gets into semantics document notes added
15:14:48 [OlivierCorby]
27 and 28 are duplicated
15:16:33 [cygri]
OlivierCorby, please try to put the name or nick of the speaker first, like this:
15:16:47 [cygri]
"davidwood: action-123 is done"
15:17:10 [AndyS]
15:17:11 [PatH]
action 123??
15:17:11 [trackbot]
Sorry, bad ACTION syntax
15:17:19 [cygri]
OlivierCorby, feel free to interrupt us if you want to ask who's speaking
15:17:58 [cygri]
PatH, sorry was just trying to give an example to the scribe
15:18:18 [PatH]
I had this vision of a *very* long action list...
15:19:17 [OlivierCorby]
Poll for F2F2
15:19:23 [davidwood]
Review the poll regarding location/dates and results:
15:19:23 [davidwood]
15:19:23 [davidwood]
15:19:56 [OlivierCorby]
19 people not responding
15:20:30 [Zakim]
15:20:35 [OlivierCorby]
towards east cost early october
15:21:06 [OlivierCorby]
please respond to the poll
15:21:24 [zwu2]
Just realized a conflict, can I change my vote?
15:21:40 [davidwood]
Please indicate attendance at:
15:21:41 [OlivierCorby]
wiki page for F2F 2 & 3
15:21:44 [sandro]
yes. zwu2
15:21:48 [gavinc]
Yes, just resubmit the form
15:21:51 [AxelPolleres]
Would there be an option to do a two site F2F with a European site connected via Video conf.?
15:21:55 [zwu2]
thanks Sandro
15:22:02 [AndyS]
+1 to Axel
15:22:07 [yvesr]
AxelPolleres, +1
15:22:09 [AxelPolleres]
15:22:13 [mischat]
+1 to Axel
15:22:19 [OlivierCorby]
Graph task force
15:22:39 [AZ]
AZ has joined #rdf-wg
15:23:02 [ivan]
ivan has joined #rdf-wg
15:23:30 [ivan]
zakim, dial ivan-voip
15:23:30 [Zakim]
ok, ivan; the call is being made
15:23:31 [Zakim]
15:23:36 [mischat]
east coast US works well re: time difference
15:23:45 [zwu2]
15:23:51 [OlivierCorby]
remote participation is possible ?
15:24:10 [LeeF]
That room at MIT is cozy :)
15:24:27 [OlivierCorby]
try with skype
15:24:43 [PatH]
I was impressed by how well the remote participation worked in the last F2F. Dont think that full video is really nfecessary.
15:24:44 [LeeF]
The nice thing about the 2-site F2F is that having multiple people at each site helps keep everyone more focused
15:24:50 [PatH]
Skype is flaky.
15:24:59 [AxelPolleres]
+! to ivan, skype not always reliable...
15:25:05 [OlivierCorby]
bad experience with skype
15:25:09 [PatH]
+1 to ivan.
15:25:24 [LeeF]
AndyS ++
15:25:35 [AxelPolleres]
+1 to Andy
15:25:48 [mischat]
15:26:09 [zwu2]
15:26:13 [AndyS]
15:26:27 [davidwood]
Zakim, who is speaking?
15:26:37 [davidwood]
ack AxerPolleres
15:26:38 [Zakim]
davidwood, listening for 10 seconds I heard sound from the following: davidwood (33%), AxelPolleres (29%), PatH (14%)
15:26:41 [Souri]
2-site F2F is very good (east coast 8-5pm is tolerable for western Europe time zone, but it does not work the other way :-))
15:26:45 [LeeF]
Everyone in the SPARQL group loved the 2-site, video-linked, F2F style.
15:26:47 [zwu2]
zakim, unmute me
15:26:47 [Zakim]
zwu2 should no longer be muted
15:26:49 [davidwood]
ack zwu
15:26:51 [gavinc]
Remote to the F2F1 as fine, other then time zone ;)
15:26:57 [davidwood]
ack AxelPolleres
15:26:58 [AxelPolleres]
zakim, ack me
15:26:59 [Zakim]
I see AndyS on the speaker queue
15:27:50 [davidwood]
ack AndyS
15:27:50 [zwu2]
zakim, mute me
15:27:51 [Zakim]
zwu2 should now be muted
15:27:55 [OlivierCorby]
Poll still open, can change vote
15:28:31 [OlivierCorby]
Site in Europe ?
15:29:33 [PatH]
mischat, good thought.
15:29:51 [mischat]
i could ask ECS, and I could ask W3C UK offices
15:29:55 [mischat]
and will report back
15:29:59 [mischat]
15:30:04 [SteveH]
mischat, ECS has the wrong system
15:30:16 [mischat]
but yes they have video geeks ...
15:30:19 [sandro]
action: mischat look into soton video conf facilities
15:30:19 [trackbot]
Created ACTION-40 - Look into soton video conf facilities [on Mischa Tuffield - due 2011-04-27].
15:30:20 [mischat]
will ask anyways
15:30:40 [mischat]
zakim, who is making noise ?
15:30:43 [AZ]
AZ has joined #rdf-wg
15:30:49 [OlivierCorby]
15:30:51 [Zakim]
mischat, listening for 10 seconds I heard sound from the following: Sandro (60%), gavinc (15%), davidwood (53%)
15:30:57 [cygri]
Topic: Skolemization
15:31:05 [OlivierCorby]
Sandro proposal from F2F
15:31:26 [OlivierCorby]
discussion on this
15:31:37 [OlivierCorby]
move on resolution ?
15:31:43 [pfps]
pfps has joined #rdf-wg
15:31:46 [cygri]
15:31:49 [cygri]
15:31:53 [davidwood]
Steve's Proposal:
15:31:53 [davidwood]
Systems wishing to skolemise bNodes, and expose those skolem constants to external systems (e.g. in query results) SHOULD mint fresh a "fresh" (globally unique) URI for each bNode.
15:31:53 [davidwood]
All systems performing skolemisation SHOULD do so in a way that they can recognise the constants once skolemised, and map back to the source bNodes where possible.
15:31:53 [davidwood]
Systems which want their skolem constants to be identifiable by other systems SHOULD use the .well-known URI prefix.
15:33:02 [OlivierCorby]
using a scheme to detect is it a bnode
15:33:09 [cygri]
SteveH: yves strongly opposed SHOULD requirement to .well-known
15:33:40 [PatH]
15:33:43 [cygri]
SteveH: tried to come up with careful wording to allow use cases
15:34:00 [yvesr]
slightly happier about this wording, although still concerned with possible over-complexity and mis-interpretations
15:34:14 [cygri]
15:34:36 [davidwood]
ack PatH
15:34:58 [cygri]
SteveH: having a URI that claims to be a bnode somewhere in the URI is nonsense
15:35:18 [cygri]
PatH: we should not publish anything that discourages use of blank nodes
15:35:50 [MacTed]
that seems incorrect, cygri... "we should not publish anything that discourages use of blank nodes" is not what I heard
15:35:51 [sandro]
pat: the only case that matters is where the publisher wants others to know that some of the URIs are special, that there is no other info to be had about this thing.
15:36:19 [cygri]
MacTed then please correct it
15:36:23 [danbri]
q+ to note the "But I didn't say that..." scenario; we shouldn't put URIs "into other's mouths"
15:36:28 [sandro]
steve: it matters for your own system to be able to recognize what was a bnode, but other systems dont need to be able to tell.
15:36:30 [pfps]
So should it be kosher for a graph store to consume something like _:b :loves :Mary and then emit something like :Mary :loves :Mary (and consider these two to be somehow close in meaning)?
15:36:34 [cygri]
15:36:39 [sandro]
pat: in that case, why do we need a standard here?
15:36:58 [MacTed]
PatH: we should not publish anything that discourages replacement of blank nodes with URIs of whatever coinage
15:37:06 [MacTed]
cygri ^^^^^
15:37:15 [MacTed]
(is what I get from his words...)
15:37:59 [gavinc]
I think it is worth the WG mentioning as folks keep inventing it on their own
15:38:05 [sandro]
steve: consequences to sparql update: if you skolemize at export time, then you have to be able to recognize uris coming in as those, so you can unify them with the bnodes in your store.
15:38:14 [MacTed]
and that's (yet another) reason to avoid bnodes
15:39:01 [sandro]
pat: we're defining a spec about publishing content. what you're talking about is a private matter.
15:39:03 [sandro]
15:39:23 [sandro]
steve: sure, it doesn't add something, but it's worth pointing out.
15:39:34 [mischat]
perhaps for the primer ?
15:39:45 [sandro]
andy: I like it as a practical experience note.
15:40:03 [ivan]
15:40:12 [PatH]
I guess Im worried that if we talk about "skolemization" and use SHOULD language, what we write gets to be holy writ.
15:40:30 [davidwood]
ack danbri
15:40:30 [Zakim]
danbri, you wanted to note the "But I didn't say that..." scenario; we shouldn't put URIs "into other's mouths"
15:40:46 [sandro]
15:40:55 [sandro]
+1 danbri for this usecase
15:41:35 [sandro]
danbri: if I take a graph from someone, skolemize, and republish, it's nice to be able to be clear that you've done this.
15:41:53 [Zakim]
15:41:59 [PatH]
danbri: yes, but that issue is, who is responsible for the 'new' URIs. And the rule surely is, whoevewr coins the URis is responsible for them.
15:42:06 [davidwood]
ack sandro
15:42:22 [danbri]
ivan, we lost you in audio?
15:42:27 [PatH]
I dont think danbri's case is a use case.
15:42:41 [danbri]
it's a mis-use case
15:42:46 [PatH]
15:43:01 [PatH]
15:43:02 [danbri]
dan loads pat's graph, does some trivial transform, republishes it, sandro consumes it, and wants to know what pat actually said
15:43:19 [SteveH]
15:43:20 [danbri]
(much as we might care to avoid muddle if dan retransmitted pat's sayings via rdf'99 reification?)
15:43:23 [davidwood]
Why is SHOULD considered holy writ? Surely MUST is...
15:43:30 [MacTed]
Zakim, unmute me
15:43:30 [Zakim]
MacTed should no longer be muted
15:43:32 [davidwood]
15:43:32 [MacTed]
15:43:39 [davidwood]
ack PatH
15:43:40 [OlivierCorby]
problem with multigraph that may share bnodes
15:44:51 [cygri]
15:45:03 [pfps]
My action-27 cannot be done until issue-12 is closed, so I've extended the due date for a while.
15:45:34 [Souri]
15:46:06 [davidwood]
ack SteveH
15:46:13 [sandro]
pat: what about just saying the one who mints an IRI is responsible for it?
15:46:27 [sandro]
sandro: yeah, that could work.
15:46:41 [webr3]
if people are talking about a specific something by name, then what is the difference between that and a uri-ref, as soon as anybody skolemizes and somebody else uses that uri, then people are talking about a specific thing rather than just something, surely?
15:46:50 [ivan_]
ivan_ has joined #rdf-wg
15:46:53 [PatH]
+1 to speaker.
15:46:56 [zwu2]
why does it need to be globally unique?
15:46:57 [sandro]
steve: I don't want to throw the baby out with the bathwater. Skolemizing is what the vast majority of stores now, and I don't like everyone ignoring the standard.
15:47:03 [pfps]
Umm, what is the standard specifying??
15:47:16 [pfps]
... that is being ignored?
15:47:16 [davidwood]
15:47:20 [ivan_]
zakim, dial ivan-voip
15:47:20 [Zakim]
ok, ivan_; the call is being made
15:47:22 [Zakim]
15:47:27 [sandro]
15:47:57 [sandro]
steve: "systems that issues URIs are responsible for them"
15:48:18 [sandro]
steve: I just think it makes sense to codify the common practice.
15:48:26 [AlexHall]
pfps, that a blank node label is not valid outside the scope of the graph in which it appears?
15:48:28 [cygri]
15:49:06 [sandro]
david: Steve, can you write a new proposal that attempts to capture that?
15:49:09 [danbri]
re responsibility, yeah i tihnk that's part of the issue
15:49:46 [webr3]
AlexHall, but there's no way you can prevent people from using that uri in another doc?
15:49:47 [sandro]
steve: "if you want your URIs to be identified as sk bno then you should..." is what Pat doesn't like
15:49:47 [danbri]
so if i ascribe to pat as author of a graph, but am actually pointing at a graph full of skolem'd bnodes i interfered with, ... is that misrepresenting pat?
15:49:58 [pfps]
I think that I would be unhappy.
15:50:00 [davidwood]
ack MacTed
15:50:14 [ivan]
15:50:51 [pfps]
15:51:06 [davidwood]
q+ to suggest we have a theory vs practice argument
15:51:07 [pfps]
15:51:37 [PatH]
For practical purposes, systems might wish to replace blank nodes by URIs. If done, the responsibility for the meaning of these newly introduced URIs lies with the publisher of the modified data.
15:51:53 [sandro]
ted: bnodes are nothing but trouble
15:52:06 [sandro]
sandro: this group isn't saying anything of the sort
15:52:17 [danbri]
minting open-ended descriptive promises to the planet - also can be troubling
15:52:20 [AndyS]
Can we have the wording for any proposal (= evolving working draft) on the wiki please - easier to point to at resolution. It took me some time (error prone?) creating a single, consolidated view.
15:52:34 [SteveH]
15:52:48 [PatH]
Whoa. this is not to do woth names being 'variable'.
15:53:02 [sandro]
+1 AndyS -- we need a stable on-wiki wording for any proposal
15:53:14 [davidwood]
ack Souri
15:53:23 [SteveH]
15:53:24 [pfps]
15:53:41 [PatH]
Cant draft text and listen at the same time.
15:53:47 [sandro]
Souri: bnodes are nothing but local-scope variables. once you make it, it's not visible outside the scope.
15:55:10 [sandro]
Souri: let the externalizer map the bnode to a URI and then it's normal, and can be used outside.
15:55:11 [PatH]
souri: agreed. Our role as a WG is not to rule on private mappings used inside tools. Its only our business when it gets published.
15:55:38 [SteveH]
15:55:41 [sandro]
Souri: ... and then provide a predicate to connect generated URI to the bnode.
15:55:44 [PatH]
souri just summarized the entire idea of skolemization, using programming terminology.
15:55:45 [Souri]
How about externalizing a bNode by specifying a mapping from bNode to a URI => _:b1 rdf:graphIRI <G1> . _:b1 rdf:bNode2IRI :someUriExternalizerChooses . (or we can use the owl:sameAs property, but that could be an overkill)
15:56:05 [cygri]
15:56:12 [PatH]
Exactly, it is up to the app. designer. NOt our business to een know about that.
15:56:14 [webr3]
fair to summarize as: skolemization should happen behind the interface before data hits the wire (so no bnodes show to the outside world)?
15:56:50 [PatH]
+1 webr3
15:57:01 [PatH]
It **is** skolemization.
15:57:06 [sandro]
ehhh, skeptical about rdf:bNode2IRI. wrong level.
15:58:16 [PatH]
Aaargh, don't say URI **represents** a bnode, please...
15:58:29 [davidwood]
15:58:30 [zwu2]
+1 to separate what and how
15:58:34 [webr3]
PatH, if everybody did that, then there would be no bnodes on the wire, as in no bnodes in serializations or in "visible" rdf ?
15:58:40 [davidwood]
ack cygri
15:58:53 [PatH]
If everybody did that, yes. BUt of course they wont ALL do it.
15:59:11 [PatH]
cygri, nooooo.
16:01:27 [sandro]
cygri: In practice you're often confronted with other people's bnodes -- by saying something about sk in our docs, ("look here's a process for when you get bnodes you didnt want...") ... I see Pat's point about it's your own private business, as long as you keep to your own URIs in the process, then why would anyone ned to know about it? I think, however, there is value in writing down the fact that it is okay to do so. It is NOT obvious. YOu have to s
16:01:27 [sandro]
pend a lot of time with RDF before you know that.
16:01:35 [davidwood]
16:01:39 [webr3]
other people skolemizing my data worries me, because for every person that does it the bnode is effectively forked, has multiple identifiers, which makes it impossible to manage, you can't merge or diff graphs fromt he same source, manage data over time etc - which completely invalidates the point of skolemizing afaict
16:02:03 [MacTed]
webr3 - that's the reason for *you* to not use bnodes...
16:02:10 [MacTed]
Zakim, mute me
16:02:11 [webr3]
MacTed, exactly
16:02:22 [Zakim]
MacTed should now be muted
16:02:24 [mischat]
i don't think people get confused between whether to use a bnode or not. If you don't want your data to be dereferencable via a URI, you use a bnode ...
16:02:30 [sandro]
cygri: This doesn't speak to making the skolem constants reconginizable. But it's nice to have a simple recipe that avoids one having to think too much.
16:03:07 [sandro]
16:03:08 [mischat]
16:03:10 [PatH]
webr3, I think that managing over time is made tricky just by there being multiple copies.
16:03:38 [zwu2]
16:03:43 [zwu2]
16:03:53 [zwu2]
zakim, unmute me
16:03:54 [danbri]
q+ to suggest a use case: dan publishes a previously private graph. it has some bnodes / skolems representing objects that dan didn't know good public URIs for, and didn't host his own. Sandro loads up that document and reading the properties of the objects it describes, figures out some good replacement URIs. Because he can see which URIs are transient bnode-derrived skolem URIs, Sandro can now REPLACE those in the graph, rather than complicate
16:03:54 [danbri]
the graph with sameAs links to the 'better' well known URIs.
16:04:00 [Souri]
16:04:04 [Zakim]
zwu2 should no longer be muted
16:04:10 [PatH]
Would like to say all this without using 'skolem' anywhere. DOn't need to revert to logic-jargon for the general reader.
16:04:16 [MacTed]
in other words -- other people will skolemize (and here's a reasonable way to do that...), when they want to reuse data sets that came to them including bnodes. That causes problems down the line. So it's best not to use bnodes.... :-)
16:04:30 [webr3]
PatH, or just by having bnodes at all, it requires the ability to say "something, let us all call it X all the time" afaict
16:04:44 [sandro]
16:04:47 [davidwood]
ack Zhe
16:05:00 [davidwood]
ack zwu
16:05:38 [sandro]
zwu2: I'm in favor of skolemizing, but isn't globally unique too strong? why not just locally unique to your store? globally unique is harder.
16:06:33 [sandro]
UUID solves this pretty well.
16:06:43 [mischat]
but then why not just use a URI
16:07:04 [PatH]
The skolem name has to be a URI. What does it mean to say that a URI is only locally unique?
16:07:08 [gavinc]
16:07:27 [gavinc]
Federated Query too
16:07:42 [AndyS]
+1 to gavinc
16:07:43 [PatH]
16:07:48 [davidwood]
ack danbri
16:07:48 [Zakim]
danbri, you wanted to suggest a use case: dan publishes a previously private graph. it has some bnodes / skolems representing objects that dan didn't know good public URIs for, and
16:07:51 [Zakim]
... didn't host his own. Sandro loads up that document and reading the properties of the objects it describes, figures out some good replacement URIs. Because he can see which URIs
16:07:56 [Zakim]
... are transient bnode-derrived skolem URIs, Sandro can now REPLACE those in the graph, rather than complicate
16:08:00 [pfps]
I think that the SHOULD should be a MUST for globally unique.
16:08:48 [cygri]
for reference, RFC 2119: SHOULD This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.
16:09:26 [davidwood]
Zakim, close the queue
16:09:27 [Zakim]
ok, davidwood, the speaker queue is closed
16:09:31 [PatH]
What danbri is talking about might be called data quality improvement. I agree, people will do this, and its a good thing. But its not skolemization.
16:09:48 [sandro]
+1 danbri: the flagging might be useful because it lets you throw away the URIs, BUT there may well be other data that grows up using it, so maybe you can't throw it away.
16:10:06 [pfps]
yes, but I think that there is a big distinction between this SHOULD and later SHOULDs in the proposal
16:10:20 [PatH]
pfps, +1
16:11:13 [PatH]
16:11:16 [davidwood]
ack Souri
16:11:18 [cygri]
sandro, in data that's all bnodes, this is literally impossible
16:11:34 [sandro]
sandro: steve, I agree the sparql-loop is useful but I don't think it's compelling, since users can just write a better query.
16:11:36 [AZ]
AZ has joined #rdf-wg
16:11:39 [SteveH]
16:11:39 [AndyS]
because the store is changing? Want to "pin" previous query results.
16:12:03 [pfps]
but then you are *not* doing skolemization, you are doing something else (which might be fine, of course)!!
16:12:03 [davidwood]
General Concept
16:12:03 [davidwood]
Systems wishing to skolemise bNodes, and expose those skolem constants to external systems (e.g. in query results) SHOULD mint fresh a "fresh" (globally unique) URI for each bNode. All systems performing skolemisation SHOULD do so in a way that they can recognise the constants once skolemised, and map back to the source bNodes where possible.
16:12:16 [cygri]
AndyS, yes
16:12:33 [sandro]
davidwood, Can we do a straw poll on the general concept to see where everyone stands?
16:12:38 [PatH]
David, delete second sentence. This is a private issue for the system, not required.
16:12:46 [danbri]
PatH, yes I wasn't arguing that the cleanup / improvement is skolem18n, but rather that it is made easier by being able to recognise which URIs are the result of skolemisation [but i've not convinced myself, since even these funny skolem'd URIs might end up popular/useful]
16:12:55 [Souri]
I feel better if "globally unique" is just a suggestion (because I think sometimes I may not need it and sometimes I may want a bNode to map to a specific URI (or two or more bNodes from different graphs to map to the same URI))
16:13:12 [webr3]
I'm frequently getting confused with this, if you have { <u> a Foo } that entails that something exists which is a Foo . So I hear lots of people saying they want a persistent name for a blanknode (as something), isn't the definition of saying that just using a URI. It sounds to me like most uses of bnodes are people saying "this thing" rather than "something", why they using bnodes?
16:13:18 [MacTed]
Zakim, unmute me
16:13:18 [Zakim]
MacTed should no longer be muted
16:13:20 [danbri]
"what you do in the privacy of your own database is your own business"
16:13:36 [Souri]
+1 to Pat
16:13:41 [AndyS]
Agree with PatH: SHOULD -> should (=we suggest)
16:13:47 [sandro]
sandro: Yes, we can weaken the second sentence.
16:14:13 [PatH]
David, dont be discouraged. We will get this done, honestly.
16:14:22 [PatH]
16:14:28 [webr3]
but why indicate it? bob a man entails that something exists that is a man - why do you even need to know "it was a bnode" because it entails all the same stuffs surely :s
16:14:31 [danbri]
i was uncomfortable with an unqualified 'should' and remain so; at one point the sentence seemed to be more qualified
16:14:36 [SteveH]
+1 to PatH
16:14:47 [mischat]
bye all
16:14:49 [Zakim]
16:14:49 [LeeF]
16:14:50 [zwu2]
16:14:50 [Zakim]
16:14:50 [Zakim]
16:14:51 [davidwood]
16:14:52 [PatH]
16:14:53 [Zakim]
16:14:54 [Zakim]
16:14:54 [Zakim]
16:14:54 [sandro]
webr3, I think there are practical engineering reasons. there are no firm logic reasons.
16:14:55 [mbrunati]
16:14:56 [Zakim]
16:14:56 [AZ]
16:14:58 [Zakim]
16:15:00 [Zakim]
16:15:02 [Zakim]
16:15:04 [Zakim]
16:15:06 [Zakim]
16:15:06 [ivan]
zakim, drop me
16:15:08 [Zakim]
16:15:10 [Zakim]
16:15:11 [AlexHall]
AlexHall has left #rdf-wg
16:15:12 [Zakim]
16:15:14 [Zakim]
Ivan is being disconnected
16:15:16 [Zakim]
16:15:18 [Zakim]
16:15:20 [Zakim]
16:15:22 [Zakim]
16:15:24 [Zakim]
16:15:26 [Zakim]
16:15:33 [AndyS]
AndyS has left #rdf-wg
16:15:38 [gavinc]
make it possible for isBlank() to do the "right" thing?
16:15:42 [Zakim]
16:15:43 [webr3]
@sandro, if there are no firm logic reasons, and the practical engineering reasons are that they want a persistent name, then where does the notion of a blank node come in to it at all?
16:15:44 [Zakim]
16:15:47 [Zakim]
16:16:28 [cygri]
webr3, because a URI may give raise to different expectations w.r.t stability etc compared to a blank node
16:16:31 [SteveH]
webr3, [butting in] mostly syntax - but it what people often use bNode syntax for as it stands
16:17:07 [sandro]
RRSAgent, make minutes public
16:17:07 [RRSAgent]
I'm logging. I don't understand 'make minutes public', sandro. Try /msg RRSAgent help
16:17:15 [Zakim]
16:17:16 [sandro]
RRSAgent, make logs public
16:17:17 [Zakim]
SW_RDFWG()11:00AM has ended
16:17:19 [Zakim]
Attendees were davidwood, Sandro, SteveH, mischat_, AxelPolleres, AndyS, gavinc, mbrunati, OlivierCorby, Scott, AZ?, AlexHall, Ivan, cygri, +34.67.92.aaaa, webr3, JFB, LeeF, zwu2,
16:17:21 [Zakim]
... +20598aabb, danbri, Souri, tomayac, FabGandon, PatH, MacTed
16:17:28 [AndyS]
AndyS has joined #rdf-wg
16:17:55 [webr3]
cygri, stability in what sense? a blank node is just saying something exists, not this thing, so there's nothing to be stable w/ a blank node ?
16:17:57 [FabGandon]
FabGandon has left #rdf-wg
16:18:07 [sandro]
I'm really seeing this as an "experimental" thing, but then tag: URIs are "experimental" as well.
16:18:29 [SteveH]
I'm seeing it more as codifying common practice
16:18:33 [cygri]
webr3, exactly. but once you skolemize it to a uri, it looks more stable but probably isnt
16:19:18 [SteveH]
URIs aren't guaranteed to be stable, it's not a promise or anything
16:19:25 [sandro]
webr3, I think the main thing about a skolem node is that you have much greater confidence you can throw away the URI. This is important if you've read the same graph many times and thus have many copies of what should have been the same triple.
16:19:47 [SteveH]
yup, that's a fair point
16:19:55 [sandro]
shall we get back on the phone? :-)
16:20:05 [SteveH]
probably wouldn't hurt
16:20:16 [cygri]
SteveH, not guaranteed of course, but minting a URI is a promise
16:20:16 [SteveH]
but I think we've lost critical mass
16:20:17 [webr3]
@sandro, in which case what was the point in skolemizing in the first place
16:20:27 [webr3]
if you're just going to throw the name away
16:20:56 [sandro]
webr3, because you wanted to pass it through something (like sparql-results-loop or graph-delta) that can't handle bnodes.
16:20:57 [SteveH]
noones making you throw it away
16:20:58 [cygri]
SteveH, if it weren't, how could you ever hyperlink to someone else's web page? (of course, some promises stronger than others)
16:21:07 [davidwood]
webr3, reading you seem to be saying −1 to a proposal that was not resolved that way. Am I missing something?
16:21:26 [SteveH]
cygri, well, it's just a best effort thing, I've got URIs from domains that I've lost control of
16:21:42 [webr3]
davidwood, perhaps I am missing something, if it wasn't resolved then no -1 from me counts :0
16:21:46 [cygri]
SteveH, a promise to best effort ;-)
16:21:57 [davidwood] is still open
16:22:12 [SteveH]
cygri, right, and best effort for a bnode skolemisation is "until I have a better idea", *shrug*
16:22:15 [cygri]
SteveH, a blank node label is not even a best effort thing. it might change anytime as far as i know
16:22:27 [SteveH]
bnoe label != skolem constant
16:23:18 [webr3]
@sandro, yes that's why i thought one would skolemize, but for the name to be useful for delta's / over time, then it has to be reliable over time, not just throw away, and in you can have multiple names for the same blank node and not be able to tell they are the same blank node, then i don't follow how that helps at all, surely it just compounds with the illussion of stability
16:23:34 [webr3]
and that illussion of stability is the whole problem that blank node identifiers introduced int he first place
16:23:45 [webr3]
so it only makes it worse afaict
16:23:55 [SteveH]
not in our experience
16:23:57 [cygri]
SteveH, if I see a bnode in your data, i know i can't link to it. if i see a URI, it's reasonable for me to expect that I can link to it. if it's a .well-known/bnode URI, then i know i probably shouldn't
16:24:14 [SteveH]
cygri, no argument from me
16:24:20 [sandro]
+1 cygri
16:24:25 [sandro]
nicely put.
16:24:29 [cygri]
webr3, see above :-)
16:24:45 [SteveH]
cygri, but people who have very stable data (inc. bNodes), and are precious of their HTTP URI space disagree
16:24:52 [sandro]
and you probably shouldn'y BECAUSE folks downstream are likely to de-skolemize and throw the sk iri away.
16:25:10 [yvesr]
cygri, relying on the shape of a URI to know whether I should link to it or not...
16:25:16 [SteveH]
sandro, likely is a bit strong, "free too" maybe
16:25:26 [cygri]
yvesr, i'm relying on an RFC, not the shape of the URI
16:25:43 [SteveH]
if the URI starts http: you should feel free to link to it, if it's not dereferencable, use tag:
16:25:43 [yvesr]
cygri, on an RFC? the one that defines well-known?
16:25:44 [SteveH]
16:25:45 [davidwood]
+1 to cygri
16:25:49 [sandro]
SteveH, can we settle on "more likely to" (than with normal IRI) ?
16:26:01 [yvesr]
cygri, there's nothing in that RFC that says i shouldn't link to a .well-known uri
16:26:05 [SteveH]
sandro, sure
16:26:19 [SteveH]
yvesr, no, quite right
16:26:24 [yvesr]
cygri, so you're relying on something built on top of it, not the RFC itself
16:26:37 [davidwood]
Updated minutes in relation to ISSUE-12 (Lee's comments). Nathan's −1 related to a proposal that was not adopted.
16:26:43 [SteveH]
if the URI starts http: you should feel free to link to it, if it's not dereferencable, use tag:
16:26:47 [SteveH]
[repost] :)
16:26:49 [sandro]
btw, SteveH I was reading about the 5 versions of UUIDs and seems like you can probably generate them cheaply enough--- 128 bits is a lot of room.
16:27:08 [webr3]
cyngri, that makes no sense to me
16:27:17 [SteveH]
sandro, yeah... but lets not dictate identifier syntax
16:27:23 [sandro]
16:27:44 [SteveH]
I suspect we (garlik) want .../$uuid/$localid
16:27:51 [SteveH]
oir something of that nature
16:27:57 [webr3]
cygri, if you want to provide a uri and say no more info to get about this thing, just don't give any more info or use a scheme that can't be looked up for more info - i don't see how that has anything to do w/ bnodes ?
16:28:38 [SteveH]
webr3, the problem comes when you do want to say more about it
16:28:52 [SteveH]
if you don't want to say anything about it, then there's no issue
16:29:28 [webr3]
but when you use a bnode you're not talking about a specific thing, you're making a general statement like "a man exists in the world" not "this specific man exists in the world"
16:32:07 [webr3]
I'm ever tempted to agree w/ pats original proposal, just loose blank nodes - I can barely see a case where people are actually using bnodes as existentials tbh, seems like peopel are just using them as a quick way of not giving soemthing a proper name (at this time) but might do later
16:33:09 [webr3]
everytime somebody see's _:b1 they think a specific think called "_:b1 in this graph only" is being talked about anyway, isn't that the problem here?
16:37:27 [sandro]
webr3, the most we can POSSIBLY do there is issue some strongly worded text about why one might want to avoid bnodes. Feel free to start drafting that text..... ?
16:38:36 [webr3]
@sandro, yup - perhaps the issue is less about skolemizing and more about avoiding blank nodes in the first place
16:39:15 [sandro]
well, for you. For me, it's about providing an intermediary service (federated query, delta) that doesn't mangle the data too badly.
16:40:18 [webr3]
@sadnro, I'd like that too.. but unsure if it's possible (when you scale up to there being two intermediaries trying to do that for the same graph and one person using both)
16:40:36 [sandro]
Agreed. Thus it's "experimental". :-
16:40:38 [sandro]
16:40:40 [webr3]
16:43:32 [webr3]
related, have been looking at whether it'd possible to do an object based rdf (like json-ld etc) which didn't have any notion of blank nodes or "anonymous objects", and it seems anonymous objects are very common - /however/ when you do diff/patch or anything over time with nested objects structures which include anonymous objects, there's no problems..
16:43:32 [webr3]
it's only when you try to break it down in to triples that you get the problems..
16:45:01 [webr3]
*although merge can still be tricky
16:45:57 [sandro]
right -- triples == merge, I think.
16:50:27 [cygri]
yvesr, W3C will register a namespace under .well-known, and in that registration it can say you shouldn't link to those URIs. or at least I can infer from the registration that those aren't stable IDs
16:53:13 [SteveH]
cygri, disagree, if you don't want people to link it, use tag:
16:53:20 [SteveH]
if it starts http: I think it should be linkable
16:53:29 [sandro]
cygri, can you put that more crisply? because obviously we want to use that URIs in more than one place for several of the use cases....
16:53:59 [cygri]
SteveH, you're right, it shouldn't say that you shouldn't link to it
16:54:30 [SteveH]
there's perfectly valid usecases for follow-your-nose on skolemised bnodes, e.g. FOAF
16:55:08 [danbri]
danbri has joined #rdf-wg
16:55:31 [sandro]
I think cygri is saying I shouldn't say eg:sandro foaf:knows <>
16:55:32 [AndyS]
UUIDs are very cheap to generate - can amortize overhead arbitrarily and do it with a integer 32 bit +1 + a 16 byte copy. OSs support epochs and clocks going backwards.
16:55:33 [cygri]
SteveH, the point is I can tell that they have been auto-generated by some process that doesn't know anything about the identity of the things identified
16:55:48 [SteveH]
cygri, yup, agreed
16:56:43 [SteveH]
cygri, but there are usecases (where the data is generated internally, and only exposed over SPARQL for e.g.) when you don't need to own up that the URIs originated from bNodes
16:56:55 [SteveH]
personally, I don't really care, but some peopel do
16:57:09 [sandro]
what do you think about that foaf:knows example? is that an open thing to do?
16:57:44 [cygri]
SteveH, sandro, in absence of any further information from the publisher, i'd expect a .well-known URI to be just as volatile as a blank node label
16:58:01 [SteveH]
cygri, fair enough
16:58:14 [sandro]
I think so too, yeah.
16:58:15 [cygri]
SteveH, sandro, but publishers can make them more stable for their internal use
16:58:28 [SteveH]
cygri, sure
16:59:54 [cygri]
so in absence of further information from the publisher i probably wouldn't link to them
17:00:30 [SteveH]
cygri, which is an argument, for people who know that their skolem constants are just as stable as "normal" URIs to not advertise the fact
17:01:02 [cygri]
SteveH, i think they'd be better off using a different namespace, that makes it look just like normal URIs
17:01:31 [cygri]
SteveH, in practice I'd hope that my store comes pre-configured with something like .well-known, but lets me override it if i want to
17:01:32 [SteveH]
cygri, yes
17:01:38 [SteveH]
cygri, yes
17:01:50 [SteveH]
17:02:19 [sandro]
sounds like a pretty good plan.
17:02:42 [cygri]
sandro, SteveH: so what document should all this go into?
17:02:49 [SteveH]
yeah, I'm not quite sure who the dissenters are
17:02:55 [SteveH]
cygri, right now, the wiki page I think
17:03:20 [sandro]
A short WG note which includes the syntax spex, I think.
17:05:29 [davidwood]
SteveH: Everyone :)
17:06:00 [SteveH]
davidwood, I'm not convinced, I think there's too much violent agreement
17:06:01 [davidwood]
cygri: +1 to good store defaults and overrides.
17:06:28 [davidwood]
Perhaps, Steve. That's why I thought we might be close to consensus for the last week.
17:06:51 [SteveH]
a note would probably be idea, /if/ the current documents don't explicitly forbid it, which i'm not clear on. by my reading they do
17:07:12 [SteveH]
but PatH said otherwise, I think
17:07:15 [cygri]
SteveH, do you think so? which parts?
17:07:35 [SteveH]
cygri, don't remember, I quoted it in email
17:08:15 [cygri]
17:08:18 [SteveH]
davidwood, I think we were very close to consensus, again :)
17:08:31 [cygri]
just had a spider running over my desk. squished it with a printout of RDF Semantics
17:08:46 [SteveH]
poor spider :-|
17:09:11 [cygri]
not a nice way to go
17:09:17 [sandro]
17:09:52 [SteveH]
sadly probably not the first or last death that will be attributable to that document
17:11:01 [cygri]
any recent suicides among the members of recent WGs?
17:11:09 [cygri]
17:12:23 [cygri]
SteveH, so the wiki currently says: "All systems performing skolemisation SHOULD do so in a way that they can recognise the constants once skolemised, and map back to the source bNodes where possible."
17:12:25 [sandro]
do you count marriages...? (never mind....)
17:13:56 [cygri]
SteveH, perhaps sufficient: "Systems may wish to perform skolemisation in a way that they can recognise ..."
17:14:40 [SteveH]
cygri, yes, that's better
17:21:52 [cygri]
should there be something like: "Systems that encounter skolem constants generated by other systems SHOULD NOT assume that the skolem URIs are permanent."
17:23:56 [mischat]
mischat has joined #rdf-wg
17:24:07 [SteveH]
cygri, no, I think that's not neccesary
17:24:56 [cygri]
SteveH, clarification, i'm talking about .well-known URIs there. still disagree?
17:25:25 [SteveH]
cygri, I don't disagree, I just don't think it adds anything
17:25:33 [SteveH]
and it may turn out not to be true
17:26:00 [SteveH]
preguessing stuff like that has been a bit of a downfall in the past
17:27:25 [cygri]
SteveH, would it be better with some language about "unless you know otherwise because the publisher makes some sort of guarantee"?
17:27:43 [SteveH]
cygri, no, I think it's best to just leave it unsaid
17:28:19 [SteveH]
and not trying to second-guess deployments
17:30:03 [SteveH]
not at all, I want to be able to use them
17:30:04 [mischat]
17:30:29 [SteveH]
I dislike having to explicitly mint URIs for everything
17:30:45 [SteveH]
the [ ... ] syntax in turtle is very handy
17:30:52 [SteveH]
but not SPARQL friendy
17:30:56 [SteveH]
17:31:55 [cygri]
SteveH sure but the URIs you're going to see for those skolemised blank nodes are likely to change each time you edit something
17:32:30 [cygri]
i'm concerned with naive users who assume that those URIs must be sort of stable (because they are URIs) and try to link to them
17:32:51 [pchampin]
pchampin has joined #rdf-wg
17:33:14 [SteveH]
17:33:19 [SteveH]
v's http:
17:33:24 [SteveH]
don't use http: if you don't mean it
17:33:27 [cygri]
SteveH, no that's a different issue
17:33:33 [SteveH]
ah, I see what you mean
17:33:40 [SteveH]
yeah, that's a concern
17:33:41 [cygri]
i mean link to in the sense of assuming they are stable names
17:33:42 [mischat]
if i had to use a URI for a blanknode, i would use the most useless uri i can think of
17:33:47 [mischat]
namely either urn or tag
17:33:59 [mischat]
as they are pretty much local variables anyways
17:34:29 [cygri]
mischat, the .well-known thing is pretty much that, just easier to register
17:34:41 [mischat]
yeah but it starts with http:
17:34:45 [mischat]
which would make me want to resolve it
17:34:51 [mischat]
but that is a different matter
17:35:17 [SteveH]
if you use http: it should really be resolvable, but that was my point when I misunderstood cygri
17:35:32 [SteveH]
instability is an issue, but it's all relative
17:36:01 [SteveH]
I'd prefer not to gaze into a crystal ball to guess what the right response is
17:36:26 [SteveH]
if we issue a note in short order, chances are we can update it with more experience down the line
17:36:36 [SteveH]
for the URIs that 4store mints, it's not been an issue that I know of
17:36:42 [cygri]
SteveH, mischat: in my experience, many users balk at funny uri schemes, that's why i prefer http://. "should be resolvable" is an issue, that's true
17:36:51 [SteveH]
but there's quite clearly not "normal" URIs, which /may/ make a difference
17:37:14 [SteveH]
machines don't care about URI schemes, users may
17:39:24 [mischat]
i see the point, but humans author the systems which write out data
17:39:42 [mischat]
but yeah, meh ... as long as bnodes dont get chucked out i am happy
17:40:59 [cygri]
mischat, SteveH: any reason why the note shouldn't list both http://...well-known AND tag:/urn:/whatever?
17:41:06 [cygri]
as options?
17:41:23 [cygri]
"Systems which want their skolem constants to be identifiable by other systems SHOULD use one of the following two options:"
17:41:38 [SteveH]
cygri, no, no reason at all
17:41:51 [SteveH]
I'd still liek to see <genid:...>, but it's maybe too optimistiic
17:42:14 [SteveH]
I still think we could say that we're aiming to register it in a note, as a none-to-subtle hint
17:42:30 [mischat]
something like, if the system decides to use a URI it is free to, but if the authors want to use an http URI it should be of form .well-known/foozle/
17:43:44 [cygri]
mischat, what's the difference in intent between your wording and the wording i gave above?
17:43:54 [mischat]
17:44:00 [cygri]
mischat ok :-)
17:44:06 [cygri]
SteveH: I would prefer <bnode:...> to be honest
17:44:09 [mischat]
trying to stress the fact that you can use any uri scheme
17:44:19 [mischat]
i mean ftp should work right ?
17:44:47 [cygri]
mischat, yeah sure, the sentence is just about the case where a system wants its URIs to be recognizable by others
17:44:51 [SteveH]
mischat, not all URI schemes support /.well-known/
17:44:53 [cygri]
if you don't care about that, use whatever you like
17:44:58 [SteveH]
17:45:02 [mischat]
17:45:41 [mischat]
my only contribution here, is that uri schemes such as tag/urn which are not resolvable are bnodes if you look at them from a certain point of view
17:45:51 [cygri]
SteveH how about <nodeid:...>? as per RDF/XML terminology?
17:46:04 [SteveH]
cygri, a nodeid is a bNode label, so NO!
17:46:16 [cygri]
SteveH, ok fair point
17:46:28 [SteveH]
c.f. bnode:
17:47:17 [cygri]
mischat, well i'd expect that a tag: uri survives reloading the document; with a blank node label, not necessarily
17:48:02 [cygri]
mischat, in other words, it's easy to make long-lived tag: uris, but hard (and sort of discouraged by the specs) with blank nodes
17:52:21 [SteveH]
"I'm looking forward to using them as graph names, actually." hehe, quite
18:02:10 [AndyS]
AndyS has joined #rdf-wg
18:17:18 [Zakim]
Zakim has left #rdf-wg
18:40:01 [pchampin]
pchampin has left #rdf-wg
18:54:51 [danbri]
danbri has joined #rdf-wg