IRC log of gld on 2013-04-12

Timestamps are in UTC.

08:07:27 [RRSAgent]
RRSAgent has joined #gld
08:07:28 [RRSAgent]
logging to
08:07:29 [trackbot]
RRSAgent, make logs world
08:07:31 [trackbot]
Zakim, this will be GLD
08:07:31 [Zakim]
ok, trackbot, I see T&S_(GLD)3:00AM already started
08:07:32 [trackbot]
Meeting: Government Linked Data Working Group Teleconference
08:07:32 [trackbot]
Date: 12 April 2013
08:08:18 [bhyland]
zakim, who is on the phone?
08:08:18 [Zakim]
On the phone I see no one
08:08:38 [MakxDekkers]
shall we disconnect and try again?
08:08:41 [martinA]
martinA has joined #gld
08:08:44 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
08:08:44 [Zakim]
ok, BartvanLeeuwen; the call is being made
08:08:55 [fadmaa]
fadmaa has joined #gld
08:09:07 [martinA]
zakim, code?
08:09:07 [Zakim]
the conference code is 4531 (tel:+1.617.761.6200, martinA
08:09:37 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
08:09:37 [Zakim]
ok, BartvanLeeuwen; the call is being made
08:09:57 [bhyland]
08:10:03 [bhyland]
or not ...
08:10:31 [MakxDekkers]
I am back in and can hear you
08:11:32 [bhyland]
zakim, who is on the call?
08:11:32 [Zakim]
On the phone I see no one
08:11:57 [cygri]
cygri has joined #gld
08:12:03 [bhyland]
zakim bye
08:12:13 [bhyland]
zakim you're dismissed
08:12:19 [martinA]
08:12:26 [bhyland]
zakim, bye
08:12:26 [Zakim]
Zakim has left #gld
08:12:35 [Zakim]
Zakim has joined #gld
08:12:50 [bhyland]
zakim, who is on the call?
08:12:51 [Zakim]
sorry, bhyland, I don't know what conference this is
08:12:51 [Zakim]
On IRC I see cygri, fadmaa, martinA, RRSAgent, HadleyBeeman, bhyland, BartvanLeeuwen, MakxDekkers, DaveReynolds, sandro, trackbot
08:12:57 [HadleyBeeman]
zakim, what conferences do you see?
08:12:58 [Zakim]
I see T&S_(GLD)3:00AM active and no others scheduled to start in the next 15 minutes
08:13:09 [MakxDekkers]
08:13:12 [cygri]
zakim, this is GLD
08:13:12 [Zakim]
ok, cygri; that matches T&S_(GLD)3:00AM
08:13:23 [cygri]
zakim, who is on the phone?
08:13:23 [Zakim]
On the phone I see no one
08:13:32 [cygri]
zakim, code?
08:13:32 [Zakim]
the conference code is 4531 (tel:+1.617.761.6200, cygri
08:13:35 [HadleyBeeman]
zakim, move GLD to here
08:13:35 [Zakim]
HadleyBeeman, this was already T&S_(GLD)3:00AM
08:13:36 [Zakim]
ok, HadleyBeeman; that matches T&S_(GLD)3:00AM
08:13:57 [HadleyBeeman]
08:14:01 [MakxDekkers]
baxk in
08:14:29 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
08:14:29 [Zakim]
ok, BartvanLeeuwen; the call is being made
08:14:57 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
08:14:57 [Zakim]
ok, BartvanLeeuwen; the call is being made
08:15:11 [HadleyBeeman]
zakim, who is on the call?
08:15:11 [Zakim]
On the phone I see no one
08:15:28 [HadleyBeeman]
phila has arrived - we're saved!
08:15:28 [MakxDekkers]
calling back in...
08:16:13 [MakxDekkers]
here again on phone
08:16:56 [MakxDekkers]
disconnecting now
08:17:04 [PhilA]
PhilA has joined #gld
08:17:12 [PhilA]
zakim, code?
08:17:12 [Zakim]
the conference code is 4531 (tel:+1.617.761.6200, PhilA
08:17:20 [PhilA]
zakim, who is here?
08:17:20 [Zakim]
On the phone I see no one
08:17:21 [Zakim]
On IRC I see PhilA, Zakim, cygri, fadmaa, martinA, RRSAgent, HadleyBeeman, bhyland, BartvanLeeuwen, MakxDekkers, DaveReynolds, sandro, trackbot
08:17:34 [PhilA]
zakim, bye
08:17:34 [Zakim]
Zakim has left #gld
08:17:36 [MakxDekkers]
yes that was what i used, same as yesterday
08:17:40 [PhilA]
zakim, this gld
08:17:52 [Zakim]
Zakim has joined #gld
08:17:56 [PhilA]
zakim, this is gld
08:17:56 [Zakim]
ok, PhilA; that matches T&S_(GLD)3:00AM
08:18:05 [PhilA]
zakim, who is here?
08:18:05 [Zakim]
On the phone I see no one
08:18:07 [Zakim]
On IRC I see PhilA, cygri, fadmaa, martinA, RRSAgent, HadleyBeeman, bhyland, BartvanLeeuwen, MakxDekkers, DaveReynolds, sandro, trackbot
08:18:20 [MakxDekkers]
calling in...
08:18:44 [PhilA]
rrsagent, make logs public
08:18:59 [MakxDekkers]
call connected
08:19:13 [MakxDekkers]
yes i can hear you
08:19:38 [PhilA]
zakim, present+ MakxDekkers, DaveReynolds, martinA
08:19:38 [Zakim]
I don't understand you, PhilA
08:19:49 [PhilA]
rrsagent, present+ MakxDekkers, DaveReynolds, martinA
08:19:49 [RRSAgent]
I'm logging. I don't understand 'present+ MakxDekkers, DaveReynolds, martinA', PhilA. Try /msg RRSAgent help
08:19:55 [HadleyBeeman]
Day 2 agenda:
08:19:56 [cygri]
present+ MakxDekkers, DaveReynolds, martinA
08:23:37 [MakxDekkers]
I am going to sign off at lunchtime today
08:24:07 [DaveReynolds]
Possible also delay CR exit until sandro is here, I know he has views on that
08:25:21 [DaveReynolds]
No thanks
08:25:36 [PhilA]
08:25:43 [PhilA]
08:27:32 [MakxDekkers]
i am very slow line so won't be able to do visual
08:27:39 [martinA]
08:27:48 [bhyland]
Ok, Deirdre is setting it up now ...
08:29:05 [PhilA]
ADMS editor's draft is at
08:29:20 [cygri]
scribe: Richard
08:29:43 [cygri]
topic: ADMS
08:30:31 [screen]
screen has joined #gld
08:34:47 [DeirdreLee]
DeirdreLee has joined #gld
08:36:47 [DaveReynolds]
One of the changes is the merge the advice for spatial and non-spatial objects.
08:37:05 [cygri]
(scribe forgot that he's supposed to scribe)
08:37:06 [DaveReynolds]
And SO will become information objects which affects http-range-14
08:37:18 [gatemezi]
gatemezi has joined #gld
08:37:20 [cygri]
PhilA: Could make ADMS a DCAT profile
08:37:36 [cygri]
…Yesterday's discussion is relevant to that, and it would work with what we resolved yesterday
08:38:02 [cygri]
… Clean and easy to say that SemanticAsset is a subclass of Dataset
08:38:03 [cygri]
08:38:35 [cygri]
… I'm not sure we need arms:distribution property, could just use dcat:distribution
08:38:43 [cygri]
08:38:50 [cygri]
… We have more properties than DCAT
08:39:07 [cygri]
… But there are very few terms that aren't either DCAT or DC Terms
08:39:22 [cygri]
bhyland: Then why does it look so complicated?
08:39:41 [cygri]
PhilA: More detail.
08:40:18 [cygri]
… adms:Identifier is perhaps the most significant addition
08:40:33 [cygri]
… Is a bit like SKOS-XL but not quite
08:40:42 [cygri]
… Allows making statements about identifiers
08:41:09 [cygri]
… Useful in RegOrg as well
08:41:44 [cygri]
… Different from Org where classification is done via skos:notation
08:42:05 [cygri]
fadmaa: One of the DCAT comments was about providing a contact point
08:42:25 [cygri]
PhilA: ADMS has contactPoint which is a VCard
08:44:10 [cygri]
fadmaa: We will probably add something for this to DCAT
08:44:16 [cygri]
… Not yet sure what namespace
08:44:38 [cygri]
PhilA: If you make dcat:contactPoint, we'd use that in ADMS
08:45:12 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
08:45:12 [Zakim]
ok, BartvanLeeuwen; the call is being made
08:45:25 [cygri]
… adms:Item is an item included in the asset
08:45:36 [bhyland]
zakim, who is on the call
08:45:36 [Zakim]
I don't understand 'who is on the call', bhyland
08:45:38 [PhilA]
zakim, who is on the call?
08:45:38 [Zakim]
On the phone I see no one
08:46:05 [MakxDekkers]
did the call drop? i am out, calling in again
08:46:21 [bhyland]
yes, the phone line dropped & we dialed back in again...
08:46:49 [MakxDekkers]
ok back in again
08:48:13 [cygri]
cygri: Instead of dcterms:hasPart, would make sense to use dcat:dataset or a subproperty
08:48:57 [DaveReynolds]
08:49:00 [DaveReynolds]
08:49:03 [DaveReynolds]
08:49:05 [cygri]
PhilA: We also have versioning between assets, xhv:next, prev, last
08:49:35 [fadmaa]
08:57:39 [fadmaa]
cygri: I don't see clear usecase for the includedItem property in ADMS seams to
08:58:02 [fadmaa]
... It'd be helpful to know what usecases need this property
08:58:26 [fadmaa]
MakxDekkers: many involved stakeholders expressed their need for such a property
08:58:45 [PhilA]
ack cygri
08:58:49 [PhilA]
ack DaveReynolds
08:59:04 [cygri]
cygri: I don't understand why adms:Item is needed. The use cases mentioned can be addressed by SKOS.
08:59:07 [fadmaa]
... I can't think of a specific usecase other than items in a codelist now
08:59:47 [cygri]
DaveReynolds: Looks like you want to use SemanticAsset as a container and you want to talk about the items. LDP is relevant here, it has a container notion
09:00:33 [cygri]
ACTION: MakxDekkers to describe use case for adms:Item and look into ldp:Container
09:00:33 [trackbot]
Error finding 'MakxDekkers'. You can review and register nicknames at <>.
09:00:57 [cygri]
ACTION: Makx to describe use case for adms:Item and look into ldp:Container
09:00:57 [trackbot]
Created ACTION-115 - Describe use case for adms:Item and look into ldp:Container [on Makx Dekkers - due 2013-04-19].
09:02:03 [cygri]
DaveReynolds: Can't you use dcterms:hasVersion instead of the xhv properties?
09:02:49 [DeirdreLee]
DeirdreLee has joined #gld
09:02:55 [cygri]
MakxDekkers: There doesn't seem to be a standard way of doing versioning
09:03:21 [cygri]
… dcterms:hasVersion doesn't give you a sequence
09:03:47 [cygri]
… You could have a network of versions related to each other
09:04:09 [cygri]
… Also a way of pointing to the last version is important
09:04:21 [cygri]
… Didn't see anything in DC Terms that works for this
09:04:58 [cygri]
DaveReynolds: Two notions being mixed here. There are versions, and then there are sequences.
09:05:15 [cygri]
… XHV not necessarily a sequence of versions, could be a sequence of other things
09:05:59 [cygri]
… Using DC Terms to say there are multiple versions, and then relating them with prev/next, would make sense
09:06:15 [cygri]
PhilA: Would a subproperty of prev/next work?
09:06:21 [cygri]
DaveReynolds: Yes.
09:07:12 [cygri]
… With DC Terms you can say here's an abstract concept, and then there are specific versions of that
09:07:37 [cygri]
09:09:19 [DaveReynolds] is the generic asset, is a specific version
09:09:28 [DaveReynolds]
maybe :)
09:11:44 [Gofran_Shukair]
Gofran_Shukair has joined #gld
09:12:29 [cygri]
cygri: For a generic versioning mechanism, it would be good to have a way of talking about the abstract, unversioned thing
09:12:38 [cygri]
MakxDekkers: ADMS doesn't do that
09:13:42 [cygri]
… next/prev can be used for things like next chapter
09:13:53 [cygri]
… So not sure whether we want a subproperty
09:14:34 [gatemezi]
gatemezi has joined #gld
09:15:24 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
09:15:24 [Zakim]
ok, BartvanLeeuwen; the call is being made
09:16:13 [cygri]
cygri: I think the use of xhv here is broken; using subproperties would be better
09:16:48 [cygri]
… Not sure how important the unversioned resource thing is. Having seen versioning modelled elsewhere, I expected to have it, but haven't thought about use cases here
09:17:08 [cygri]
PhilA: We should say that ADMS is not designed to do that
09:17:55 [cygri]
… So we should define subproperties of prev/next/last, and say that ADMS is not a generic versioning mechanism
09:18:24 [cygri]
MakxDekkers: Someone should define a generic versioning mechanism, it's needed and many people do it wrong
09:18:30 [DaveReynolds]
FWIW in the UK have a tiny version: vocab. Illustrated at top of
09:18:38 [cygri]
… Uneasy about including more ADMS-specific stuff for this
09:19:08 [DaveReynolds]
09:19:11 [DaveReynolds]
09:19:14 [DaveReynolds]
09:20:31 [MakxDekkers]
can't realy hear
09:20:40 [cygri]
fadmaa: Good to have ADMS as a DCAT profile. Makes sense.
09:20:57 [cygri]
09:21:03 [bhyland1]
bhyland1 has joined #gld
09:21:34 [cygri]
… In DCAT, we specifically left open lots of ranges, to be restricted in profiles. ADMS does that in some places
09:22:24 [cygri]
… the PeriodOfTime modelling is in the diagram but seems to be missing from the spec
09:22:44 [fadmaa]
ack me
09:23:49 [cygri]
ACTION: PhilA to update ADMS to define subproperties for prev/next/last and write text around that
09:23:49 [trackbot]
Created ACTION-116 - Update ADMS to define subproperties for prev/next/last and write text around that [on Phil Archer - due 2013-04-19].
09:24:42 [cygri]
ACTION: PhilA to update ADMS to use DCAT properties instead of adms:distribution and dct:hasPart
09:24:42 [trackbot]
Created ACTION-117 - Update ADMS to use DCAT properties instead of adms:distribution and dct:hasPart [on Phil Archer - due 2013-04-19].
09:27:15 [Gofran_Shukair]
I am back to the call again
09:27:31 [cygri]
(discussion on whether DCAT catalog->dataset is hasPart or not)
09:27:42 [cygri]
fadmaa: (going through ADMS comments)
09:28:36 [Gofran_Shukair]
I have addressed 1. All typos reported were fixed 2. Clarifying the first part of the introduction that defines semantic asset 3. Move the description of the original development of ADMS to a new acknowledgments section
09:28:37 [cygri]
… James suggested to use foaf:page instead of adms:relatedWebPage
09:29:55 [cygri]
… James suggested to rename representationTechnique, although no proposal
09:30:13 [cygri]
PhilA: It's horrible, but already implemented, and if no one can think of anything better we should just leave it
09:31:31 [DaveReynolds]
Presumably could publish as an updated WD even if not ready to go to Note.
09:31:48 [fadmaa]
PhilA: We have to wait for a couple of issues on DCAT to be resolved before being able to come to the group with a ready version of the ADMS as a note
09:32:25 [fadmaa]
cygri: Few issues on DCAT that could affect ADMS might take some time to be resolved
09:32:31 [MakxDekkers]
better to wait and publishe a note that is consistent with final DCAT
09:33:15 [fadmaa]
... ADMS can wait till DCAT is ready but I think it is better to publish it as it is now and then update it when DCAT is ready
09:33:28 [MakxDekkers]
as long as it is versioned ;-)
09:33:53 [DaveReynolds]
Sandro yesterday recommended for BP that if there's going to be multple versions then publish as WD (with status saying it would be a note)
09:34:36 [MakxDekkers]
agree with Dave WD until DCAT is ready
09:34:38 [cygri]
PhilA: Could either publish ADMS Note now, and update it when DCAT is ready; or wait until DCAT is done and only publish ADMS then
09:35:19 [cygri]
HadleyBeeman: My preference is to get something out now
09:35:26 [cygri]
cygri: Agree with HadleyBeeman
09:36:01 [Gofran_Shukair]
09:38:06 [DaveReynolds]
Is you want a WG vote for a Note, as opposed to WD, then we'll need adequate time to review. That's not likely to be possible for next Thursday.
09:38:07 [cygri]
ACTION: cygri to raise issue on adms:includedAssed vs adms:includedItem
09:38:07 [trackbot]
Created ACTION-118 - Raise issue on adms:includedAssed vs adms:includedItem [on Richard Cyganiak - due 2013-04-19].
09:38:30 [cygri]
PhilA: Reasonable to take this comment back to ISA
09:38:51 [DaveReynolds]
09:38:58 [cygri]
ack me
09:40:05 [cygri]
DaveReynolds: A draft should be available ahead of date where vote will be taken so that WG members can review it
09:40:48 [cygri]
PhilA: Notice had been given
09:41:06 [cygri]
… And things that will still be updated have been discussed today
09:41:23 [Gofran_Shukair]
I have to le
09:41:32 [Gofran_Shukair]
I have to leave now sorry
09:41:45 [Gofran_Shukair]
talk to you later
09:42:00 [Gofran_Shukair]
09:42:15 [Gofran_Shukair]
thanks to you Phil
09:42:25 [cygri]
topic: Organization Vocabulary
09:43:34 [DaveReynolds]
09:43:37 [HadleyBeeman]
scribe: hadleybeeman
09:44:15 [HadleyBeeman]
DaveReynolds: From original last call for ORG, we had a number of comments which we've addressed.
09:44:42 [HadleyBeeman]
… Since then, we've had this accidentally extended period for last call, we've had 2 more comments.
09:45:09 [HadleyBeeman]
… Resolved Bill Roberts's comment, he is satisfied.
09:45:28 [HadleyBeeman]
… All comments have been addressed, we have evidence of that.
09:45:40 [cygri]
DaveReynolds, our call dropped
09:45:42 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
09:45:42 [Zakim]
ok, BartvanLeeuwen; the call is being made
09:46:58 [HadleyBeeman]
DaveReynolds: Question: whether the changes we made after the first last call were substantive, and should we have done another last call? The document was edited, but the second last call pointed to the original document without those changes in.
09:47:47 [HadleyBeeman]
… The changes we made, I would argue, wouldn't have invalidated existing implementations.
09:47:59 [HadleyBeeman]
… Assuming we're happy with that, and the transition meeting will be happy with that, we're fine.
09:49:09 [HadleyBeeman]
… One more comment from Joao-Paulo says that the diagram isn't normative, doesn't cover all features of ORG; he'd prefer it to be complete and in proper UML style. I prefer the current diagram, but I won't object if someone wants to provide that. No response. This is purely an editorial thing, but we need to decide to have one or decide we don't need one.
09:49:18 [cygri]
09:49:27 [DaveReynolds]
09:49:30 [HadleyBeeman]
… Next step: put things together for a transition meeting.
09:50:22 [BartvanLeeuwen]
09:50:40 [HadleyBeeman]
Cygri: Comment on the diagram (general comment on all our vocabulary documents). IT would be terrific if in those diagrams, I could click on things and go to the definition of the thing that was clicked. I understand this is difficult to do. Do we as a working group want to do this for all our specs?
09:51:00 [HadleyBeeman]
PhilA: clickable schema leading to an SVG diagram? that would be great.
09:51:30 [HadleyBeeman]
Bartvanleeuwen: All the vocabularies have different types of diagrams. I agree, but I'm not sure we should take our time right now.
09:51:46 [HadleyBeeman]
cygri: It is purely an editorial comment. Could be updated in the future.
09:51:51 [BartvanLeeuwen]
ack me
09:52:29 [cygri]
ack me
09:53:10 [HadleyBeeman]
davereynolds: It would be great. It would just take time to do it.
09:53:55 [MakxDekkers]
yes I
09:54:09 [HadleyBeeman]
Hadleybeeman: How is this not addressed by other working groups?
09:54:14 [MakxDekkers]
sorry slip of keyboard
09:54:44 [HadleyBeeman]
cygri: This is something we could raise on the mailing list.
09:55:50 [HadleyBeeman]
davereynolds: there was a version of RDF Gravity that could visualise our ontologies. But the clickable diagram and low cost requirements are challenging.
09:57:58 [DaveReynolds]
09:59:15 [HadleyBeeman]
DaveReynolds: If we're happy that we don't need another Last Call, then we just need to packing things up for transition to CR. Should take a few weeks to pull together that documentation (I'm guessing). I'm tied up next week, so two weeks after puts it into early may. Then a few weeks to transition/publication.
10:00:49 [HadleyBeeman]
PROPOSED: having examined the summary for ORG comments, another iteration of Last Call for ORG is not needed.
10:01:00 [PhilA]
10:01:01 [DaveReynolds]
10:01:02 [bhyland1]
10:01:03 [fadmaa]
10:01:05 [gatemezi]
10:01:06 [BartvanLeeuwen]
10:01:15 [HadleyBeeman]
RESOLVED: having examined the summary for ORG comments, another iteration of Last Call for ORG is not needed.
10:01:18 [DeirdreLee]
10:01:23 [MakxDekkers]
10:01:30 [martinA]
10:02:55 [HadleyBeeman]
Sandro: Someone should review the editorial changes to ORG before we vote to go to CR
10:03:02 [HadleyBeeman]
… (two people)
10:03:23 [cygri]
10:03:29 [bhyland1]
zakim, who is on the call?
10:03:29 [Zakim]
On the phone I see no one
10:04:22 [HadleyBeeman]
Sandro: review could happen today?
10:05:04 [PhilA]
zakim, mute martinA
10:05:04 [Zakim]
sorry, PhilA, I do not know which phone connection belongs to martinA
10:06:14 [HadleyBeeman]
DaveReynolds: Last change to ORG was in response to the response to the PROV WG. Has anyone looked at it since then?
10:06:27 [HadleyBeeman]
… Could we have a volunteer to do that?
10:06:37 [HadleyBeeman]
… We can then vote when that's done and we have the exit criteria for CR.
10:06:38 [fadmaa]
I can do that
10:06:55 [cygri]
me too
10:08:00 [HadleyBeeman]
5 min break
10:08:21 [bhyland1]
Resuming at 11.15AM Dublin time
10:15:55 [MakxDekkers]
i am back in
10:16:00 [martinA]
10:19:09 [HadleyBeeman]
10:19:22 [HadleyBeeman]
Topic: RegORG
10:19:38 [HadleyBeeman]
PhilA: RegORG, we've previously decided, is a profile of ORG.
10:20:06 [HadleyBeeman]
… We have no open Tracker issues on RegORG
10:20:09 [Zakim]
Zakim has left #gld
10:20:15 [Zakim]
Zakim has joined #gld
10:20:50 [sandro]
10:21:06 [BartvanLeeuwen]
10:21:41 [DaveReynolds]
10:21:49 [HadleyBeeman]
… I need to reconcile dcterms:type as aproplert on adms:identifier, or should it just be a type:literal on adms:notation. (This is a new issue on regORG: alignment on ADMS identifier)
10:22:55 [HadleyBeeman]
davereynolds: noticed admsscheme:agency ?
10:23:07 [HadleyBeeman]
phila: It's a literal. I need to make that clearer.
10:23:33 [HadleyBeeman]
10:24:08 [HadleyBeeman]
phila: best practice doesn't say anything about encoding skos?
10:24:53 [HadleyBeeman]
DaveReynolds: SDMS attribute code. Published as XML, using Jena. Mechanical translation. Info may be in the google group.
10:25:01 [martinA]
martinA has left #gld
10:25:05 [HadleyBeeman]
PhilA: that is an example on how to do it; would be nice to have it as documentation.
10:25:51 [HadleyBeeman]
… I'd like to do something on how to create a SKOS concept scheme from a PDF of a vocabulary. We haven't got time to do it though.
10:26:00 [martinA]
martinA has joined #gld
10:26:19 [HadleyBeeman]
bhyland: I'm making a note to see about addressing this in Best Practices. People always ask how to do that.
10:26:54 [HadleyBeeman]
philA: in mapping between ORG and RegORG: it should be a vertical list (to be mobile-friendly). I'll get rid of the table.
10:28:04 [HadleyBeeman]
… Want to point from org:organization to the registeredOrganization. We should just keep the domain of registeredOrganization as foaf:agent
10:28:36 [HadleyBeeman]
… Marios, as co-editor, will be helping with this.
10:29:38 [HadleyBeeman]
… I will fix the issue re adms:identifier. Next version to group, which will be the final note. Bottom line, nothing huge on it. We haven't received many comments (though we may not have asked them for it.)
10:31:10 [HadleyBeeman]
scribe: deirdrelee
10:33:05 [bhyland1]
Topic: Exit Criteria for CR
10:33:20 [DeirdreLee]
sandro: requirement that at least 2 systems have to pass the test suite is to prove that the spec could be implemented and could be implemented interoperably
10:33:20 [HadleyBeeman]
Each item in a test suite needs to be passed by two implementations
10:34:44 [DeirdreLee]
... if there is no software consuming the data, it is difficult to identify if data is being published incorrectly
10:35:01 [DeirdreLee]
... therefore there's a need to show data consumption as well as publication
10:35:11 [PhilA]
10:35:34 [DeirdreLee]
... should every term be used? I would say so, as that is the only way we can show that the term is being correctly utilised
10:36:06 [DeirdreLee]
... so every term in spec should be used in 2 publication implementations, and most terms in 2 consumption implementations
10:36:17 [bhyland1]
10:36:20 [HadleyBeeman]
10:36:27 [bhyland1]
10:36:30 [bhyland1]
10:36:38 [PhilA]
ack DaveReynolds
10:37:15 [bhyland1]
ack bhyland
10:37:27 [DeirdreLee]
DaveReynolds: agrees that this is a high bar. it would require people on our side to record all implementations and to see if every term is being used from every vocab
10:37:55 [DeirdreLee]
... to show that we pass the exit criteria test
10:39:06 [MakxDekkers]
I wonder whether things like dcterms and foaf would pass the test?
10:39:24 [DeirdreLee]
... if consumption is to prove that terms are being interpreted in the same way, we would need a consumption implementation from a different publisher implementer
10:40:06 [DeirdreLee]
... another point: if we have no evidence of a given term being used, should the term be deleted?
10:40:15 [DeirdreLee]
... DaveReynolds does not think so
10:40:31 [DeirdreLee]
... it may be useful in the future
10:40:45 [DeirdreLee]
sandro: we could argue for a lower bar
10:40:55 [DaveReynolds]
10:40:55 [PhilA]
ack me
10:41:03 [bhyland1]
10:41:04 [DeirdreLee]
DaveReynolds: difficulty with a lower bar is that it is more difficult to define
10:42:05 [DeirdreLee]
PhilA: PhilA has done some work iwht the Greek gov using google refine, suggesting it could be plugged into the organogram
10:42:32 [DeirdreLee]
... organogram from
10:42:44 [BartvanLeeuwen]
10:43:20 [sandro]
+1 phil Yeah, it's probably enough to have multiple instance data files and inspect them by hand to make sure the terms are being used consistently
10:43:40 [DeirdreLee]
... so for org ontology, PhilA proposes....?
10:44:07 [DeirdreLee]
... for dcat, we could look at to see how they're using it
10:44:21 [MakxDekkers]
daon't forget the spanish DCAT profile
10:44:34 [cygri]
10:44:53 [PhilA]
ack HadleyBeeman
10:45:04 [HadleyBeeman]
10:45:11 [DeirdreLee]
... for dcat, we have a wide range of people looking at it & implementations,which we can use to see how it is being used, how the terms are being understood
10:45:41 [martinA]
Not for the English version yet
10:46:05 [DaveReynolds]
PROV did a very high bar process
10:46:07 [HadleyBeeman]
ack me
10:46:08 [DeirdreLee]
HadleyBeeman: are there any other groups that have gone through this CR process, who we can learn from
10:46:15 [DeirdreLee]
cygri: skos & prov
10:46:27 [DeirdreLee]
DaveReynolds: sparql
10:46:33 [PhilA]
Prov CR Exit Criteria
10:46:33 [martinA]
technical specification on PSI reuse in Spain:
10:47:03 [gatemezi]
Some uses of ORG:
10:47:09 [DeirdreLee]
bhyland1: would question if existing rec would reach the high bar
10:47:23 [PhilA]
For the record is a profile of DCAT as used in Spain (doc is in Spanish)
10:47:43 [DeirdreLee]
sandro: happy to lower bar to what DaveReynolds suggested, that implementations could be manually inspected
10:47:46 [HadleyBeeman]
for reference, the exit criteria for SKOS to get to CR: 1. At least two implementations have been demonstrated that use
10:47:46 [HadleyBeeman]
features of the SKOS vocabulary. Other vocabularies that use
10:47:46 [HadleyBeeman]
SKOS are candidates for inclusion in the implementation report.
10:47:46 [HadleyBeeman]
2. All issues raised during the CR period against this document
10:47:46 [HadleyBeeman]
have received formal responses.
10:47:51 [BartvanLeeuwen]
ack bhyland
10:47:59 [MakxDekkers]
the European AP draft is at
10:48:27 [DeirdreLee]
sandro: this is not uncommon, WEBIDL has been stuck in CR for over a year, although it is widely used
10:48:40 [martinA]
Also for the record, this is a draft with a guide for using the Spanish DCAT Profile and other best practices on PSI publication using Linked Data: (also in Spanish, I'm sorry)
10:49:10 [DeirdreLee]
HadleyBeeman: the skos exit criteria is much less laborious
10:49:24 [DeirdreLee]
PhilA: prov-o exit criteria is very complex
10:49:36 [cygri]
10:49:51 [DeirdreLee]
HadleyBeeman: have we settled on 2 implementations?
10:50:08 [DaveReynolds]
10:50:57 [DeirdreLee]
PhilA: we should at least have one publication implementation for every term, and if a term is used in multiple implementations, it should be used in a common way
10:51:01 [cygri]
10:51:24 [HadleyBeeman]
sandro: do we need example files? Not a full test suite, but evidence it works?
10:52:04 [DeirdreLee]
BartvanLeeuwen: is publication of org data for the community directory count as a valid implementation (using or not using diromatic)
10:52:21 [DeirdreLee]
bhyland1: yes
10:53:37 [BartvanLeeuwen]
ack me
10:53:42 [DeirdreLee]
PhilA: we could use the Community Directory as an implementation
10:54:57 [DeirdreLee]
BartvanLeeuwen: could we describe the comm directory using dcat?
10:55:02 [DeirdreLee]
fadmaa: comm directory is not a data catalogue
10:55:14 [MakxDekkers]
anything can be a dataset
10:55:26 [cygri]
10:55:42 [DeirdreLee]
PhilA: comm directory could be a data catalogue
10:55:58 [gatemezi]
An attempt of using DCAT by the city of Montpellier..
10:56:11 [gatemezi]
They are still using an old version of DCAT
10:57:00 [martinA] (Catalonia)
10:57:22 [sandro]
10:57:25 [PhilA]
ack cygri
10:57:52 [gatemezi]
10:58:47 [DeirdreLee]
cygri: calling a publisher an implemtation of a vocab is not a very useful way of showing that a vocab is right
10:59:10 [DeirdreLee]
... anyone can throw something together, but this does not show that this facilitates successful interoperability
10:59:48 [bhyland1]
+10 to a test suite that consumes a given vocabulary and does something basic & simple.
11:00:09 [DeirdreLee]
... a better way might be to create a test suite that uses the vocabulary, and show that there are two consumers can do something useful with these 'agreed upon' published data
11:00:12 [PhilA]
q+ with possible route to 'implementations'
11:00:12 [martinA]
I'm sorry, I have to leave. Enjoy the rest of the meeting. I'll read the minutes. See you soon.
11:00:16 [bhyland1]
That would help vocab designers, publishers & consumers.
11:00:24 [PhilA]
q+ to offer a possible route to 'implementations'
11:00:37 [sandro]
q+ to ask about grad students and interns
11:00:44 [martinA]
11:00:45 [DeirdreLee]
... this way we have a testing ecosystem and would be much stronger evidence that the vocabs are interoperable
11:00:48 [martinA]
martinA has left #gld
11:01:04 [bhyland1]
FTR - I'm much more comfortable with what cygri just decribed.
11:01:37 [DeirdreLee]
bhyland1: much more consistent approach, it compares apples with apples
11:02:48 [DaveReynolds]
11:03:07 [DeirdreLee]
... if a simple client can use the data, as a simple visualisation, listing or whatever, it would be more consistent
11:03:27 [DeirdreLee]
Sandro: isn't this bar higher than what was initially proposed?
11:04:13 [sandro]
11:04:21 [PhilA]
11:04:35 [DeirdreLee]
bhyland1: to have something like an RDF validator, this being a vocab validator, would prove usefulness of vocab
11:05:04 [DeirdreLee]
... simply publishing data using a vocab does not show the usefulness or the interoperability of the vocab
11:05:53 [DeirdreLee]
DaveReynolds: we do not have the option of automatic vocab validation
11:06:10 [bhyland1]
This is a suggestion for future vocabulary efforts to validate fitness for use.
11:06:11 [cygri]
(I agree with DaveReynolds, automatic validation is not really possible here. appropriate use can't be automatically validated.)
11:06:13 [sandro]
DaveReynolds: We could validate at some low level, like types, but we can't automatically tell whether this person is really the manager vs subordinate
11:06:15 [DeirdreLee]
.. but validation of apropriateness of use would take manual inspection
11:06:26 [cygri]
11:06:40 [DeirdreLee]
ack DaveReynolds
11:07:11 [DeirdreLee]
... a visualisation of data would also not be a validation
11:07:37 [DeirdreLee]
... therefore we should define what we mean by 'consumption' for the CR exit criteria
11:08:32 [sandro]
q+ sandro2 to say I think visualization is okay if it's semantic (like an organogram)
11:08:43 [DeirdreLee]
PhilA: to clarify what DaveReynolds said, a tool can be built to visualise data modelled by a vocab, but it does not demonstrate that the vocab and its terms have been understood and used consistently
11:09:17 [HadleyBeeman]
11:09:21 [PhilA]
q+ to look at the conformance criteria such as
11:09:42 [DeirdreLee]
bhyland1: if there is a sample of test publication data, there will be consistency with the consumption approach
11:09:51 [gatemezi]
Yes PhillA.. a conformance-validator vocab
11:10:05 [sandro]
(I'm so confused at bhyland1 switching from bar-1 is too high and bar-2 is great, when bar-2 is higher than bar-1)
11:10:11 [DeirdreLee]
... the point is to show data interoperability
11:10:12 [gatemezi]
11:10:31 [sandro]
ack sandro
11:10:31 [Zakim]
sandro, you wanted to ask about grad students and interns
11:10:36 [fadmaa]
scribe: fadmaa
11:11:03 [fadmaa]
sandro: cygri's bar seams to be higher than the one I set
11:11:18 [DaveReynolds]
11:11:34 [bhyland1]
Yes Sandro, I agree that the vocab validator in a way is higher than what you described, but it responds to the suggestion that I understood Richard was making about a test harness.
11:11:41 [fadmaa]
sandro: would you consider having a student or an intern working on a project as an implementation
11:12:06 [sandro]
11:12:09 [sandro]
fadmaa, yes,
11:12:12 [fadmaa]
cygri: I am not so interested in who is using these vocabularies because you can use it but you might be using it wrongly
11:12:22 [bhyland1]
cyrgi: A developer could "use" a vocab but do it all wrong. Thus, use alone doesn't mean the vocab is "good".
11:13:48 [fadmaa]
cygri: the testcase can be something like I have an organization and it contains a department named X. if this is expressed in triples using org
11:13:49 [sandro]
imagining "org-check" which checks for things like loops in the hierarchy
11:14:25 [fadmaa]
... if a consumer can understand the existence of an organization and department by looking into the triples that can be called a success
11:14:32 [fadmaa]
... this doesn't have to be automatic
11:15:13 [sandro]
q- sandro2
11:15:18 [sandro]
ack cygri
11:15:30 [bhyland1]
FTR - I wasn't suggesting painting a picture of class names & properties as a harness. That would be useless IMO.
11:15:39 [fadmaa]
cygri: regarding the issue of consuming the triples via some visualizations of the data, this is actually just another way of presenting the triple and not really understanding the vocabulary
11:15:53 [PhilA]
ack me
11:15:53 [Zakim]
PhilA, you wanted to look at the conformance criteria such as
11:16:26 [fadmaa]
PhilA: we should look into the conformance sections we added to all the vocabularies
11:16:56 [fadmaa]
... the conformance states that you use the proper term from the vocabulary when one exists
11:17:05 [gatemezi]
s/PhilA: /PhilA,/
11:17:11 [fadmaa]
... it doesn't have to be a use of the whole vocabulary
11:17:42 [fadmaa]
... we can use the conformance as exit criteria
11:18:12 [fadmaa]
... I did some work on the org vocabulary
11:18:36 [HadleyBeeman]
11:18:37 [MakxDekkers]
had that discussion at DCMI; the questions was: can you conform to Dublin Core if you use only one of the terms?
11:18:48 [fadmaa]
... in the spirit of using data represented in the ORG vocabulary based on the conformance criteria
11:20:05 [PhilA]
The pilot study on ORG done under the ISA Programme is at
11:20:32 [fadmaa]
DaveReynolds, It'll be very hard to write some criteria of correct consumption of a vocabulary
11:21:17 [fadmaa]
DaveReynolds, what is the criteria of how independent the implementations should be
11:21:32 [fadmaa]
sandro, there is no restriction on this
11:21:34 [cygri]
11:21:45 [DaveReynolds]
ack me
11:21:56 [HadleyBeeman]
11:22:03 [cygri]
q+ to respond to DaveReynolds' point about generic RDF consumers
11:22:34 [HadleyBeeman]
ack cygri
11:22:34 [Zakim]
cygri, you wanted to respond to DaveReynolds' point about generic RDF consumers
11:22:36 [fadmaa]
cygri: negative test cases is the answer to the generic RDF consumer
11:23:17 [fadmaa]
... a conforming implementation shouldn't use terms from outside the vocabulary when one exists in it
11:23:29 [DaveReynolds]
11:24:08 [fadmaa]
... an ORG implementation will consume non-conformant data differently than the conformant data
11:24:30 [PhilA]
ack HadleyBeeman
11:25:16 [PhilA]
HadleyBeeman: It's important for us to be able to demonstrate that consuming implementations are useful. But we need to show that the vocab is useful. is ORG useful?
11:25:21 [fadmaa]
HadleyBeeman: it is important to be able to demonstrate that what we have done is useful but am worried about putting too much emphasis about the existence of implementation
11:25:36 [PhilA]
ack DaveReynolds
11:26:32 [HadleyBeeman]
Yes. Also that there are reasons that might stop implementations consuming ORG, for example, that might not mean that ORG is useful and well-constructed. (Market activities, etc.) There are other dependencies that are not just "have we built ORG well and usefully?"
11:26:43 [fadmaa]
DaveReynolds: vocabularies are extensible. if the vocabulary was extended by subclassing for example an implementation would pass the testcase only if it does RDFS inference
11:27:38 [fadmaa]
DaveReynolds: Cube can be mechanically processed while ORG and DCAT are more targeted for visualizations and the like
11:28:07 [fadmaa]
... criteria for cube might need to be different than those for ORG and DCAT
11:28:20 [fadmaa]
... we have a checking criteria for Cube
11:29:15 [fadmaa]
sandro: we have two opinions regarding the exit criteria
11:29:42 [HadleyBeeman]
sandro: the question seems to be: do we need code written to consume the vocabularies, or not?
11:30:43 [DaveReynolds]
DaveReynolds: QB includes explicit integrity constraints. The exit criteria may need to include validation of publications against those constraints.
11:30:50 [fadmaa]
PhilA: reading the conformance section of Data Cube, you still need a human being to confirm conformance
11:31:25 [fadmaa]
sandro: the RDF validator checks part automatically but asks human input for other parts
11:31:58 [fadmaa]
PhilA: a tool is useful but you still need a human judgement
11:32:06 [bhyland1]
11:32:12 [MakxDekkers]
11:32:13 [fadmaa]
... it'd be great to have a tool but it is not mandatory
11:32:18 [fadmaa]
sandro: disagree
11:32:52 [fadmaa]
11:33:08 [sandro]
PROPOSED-1: Our Exit Criteria for our Vocabs is: each term in the vocab used in at least two government data sources and two humans have inspected each data source and confirmed they are used in conformance to the spec.
11:33:50 [MakxDekkers]
11:34:05 [PhilA]
ack MakxDekkers
11:34:12 [sandro]
PROPOSED: Our Exit Criteria for our Vocabs is: each term in the vocab used in at least two data sources (ideally governments) and two humans have inspected each data source and confirmed they are used in conformance to the spec.
11:34:42 [fadmaa]
MakxDekkers: what about the terms we are reusing from other vocabularies
11:34:46 [cygri]
"each term whose use is recommended in the spec"
11:34:46 [HadleyBeeman]
I like that: "ideally government". Keeps us out of the grey areas of who is government and who isn't, and what to do with suppliers hosting on behalf of government, etc.
11:34:55 [fadmaa]
... e.g. the DCTerms reused in DCAT
11:35:02 [sandro]
MakxDekkers:Are we only looking at dcat:* or also dc:* stuff that's in DCAT?
11:35:17 [fadmaa]
PhilA: we are looking at all of the properties
11:35:18 [sandro]
PhilA: I think we're looking at all of them. Everything used in the diagram.
11:35:37 [sandro]
PROPOSED: Our Exit Criteria for our Vocabs is: each term in the vocab used in at least two data sources and two humans have inspected each data source and confirmed they are used in conformance to the spec.
11:35:52 [fadmaa]
11:36:00 [PhilA]
ack fadmaa
11:36:19 [PhilA]
fadmaa: What if someone is using the vocabulary in the wrong way?
11:36:48 [cygri]
11:37:25 [fadmaa]
sandro: we also have to report on all implementations we are aware of
11:37:29 [PhilA]
ack cygri
11:37:33 [HadleyBeeman]
11:37:37 [bhyland1]
abstain - I think use of each term in 2 different data sources remains a very difficult bar.
11:37:53 [sandro]
If you don't like this bar, but can live with it, vote "-0" or "0"
11:38:12 [fadmaa]
cygri: I don't like checking conformance by having two humans checking for it
11:38:46 [fadmaa]
... would it be any two humans?
11:39:01 [fadmaa]
sandro: do you want to restrict this to members of the WG?
11:39:06 [fadmaa]
cygri: maybe
11:39:30 [fadmaa]
cygri: but I think we need to do the possible automatic checking
11:39:53 [fadmaa]
cygri: this can be done on both the consuming and publishing sides
11:40:08 [fadmaa]
... I can write a SPARQL query to check a producer implementation
11:40:16 [fadmaa]
... for example
11:40:42 [PhilA]
q+ To support SPARQL
11:40:51 [sandro]
PROPOSED: Our Exit Criteria for our Vocabs is: each term in the vocab used in at least two data sources and two people (selected by the WG for their expertise) have inspected each data source and confirmed they are used in conformance to the spec. Inspection should use appropriate software tools to assist in their work.
11:41:44 [sandro]
this doesn't say whether we will or wont make a test suite.
11:41:55 [bhyland1]
11:41:56 [sandro]
exit criteria is not based on test suite, correct.
11:42:13 [sandro]
I hear cygri volunteering to develop a test suite and tools.....
11:42:17 [PhilA]
ack HadleyBeeman
11:42:40 [fadmaa]
HadleyBeeman: asking sandro about reporting on all implementations we find
11:42:51 [fadmaa]
... even the non-compliant ones
11:42:58 [fadmaa]
sandro: I think so
11:43:07 [DaveReynolds]
11:43:10 [PhilA]
ack me
11:43:10 [Zakim]
PhilA, you wanted to support SPARQL
11:43:42 [sandro]
sandro: The Transition Request includes an Implementation Report (describing what we know about all implementations)
11:44:07 [fadmaa]
PhilA: I agree that there are things we can test and provide results reports. This can be done via SPARQL queries and doesn't necessarily require building tools
11:44:31 [fadmaa]
PhilA: and we publish the SPARQL queries so human reviewers can use that
11:45:11 [MakxDekkers]
by the way, my time is running out.
11:45:35 [DaveReynolds]
11:46:01 [sandro]
q+ to say yes there is time to build tools
11:46:06 [sandro]
ack bhyland1
11:46:12 [DaveReynolds]
11:46:15 [fadmaa]
bhyland1: practically, given the time restriction, we can't build tools
11:46:28 [HadleyBeeman]
ack sandro
11:46:28 [Zakim]
sandro, you wanted to say yes there is time to build tools
11:46:34 [PhilA]
PROPOSED: Our Exit Criteria for our Vocabs is: each term in the vocab used in at least two data sources and two people (selected by the WG for their expertise) have inspected each data source and confirmed they are used in conformance to the spec. Inspection should use appropriate software tools to assist in their work and will include one or more SPARQL queries that, in the WG's view, SHOULD return useful data from conformant data. Reporting will
11:46:34 [PhilA]
include any data that doesn't conform as well as that which does.
11:46:54 [fadmaa]
... the best we can do is have people identify implementations and find a practical way to define exit criteria and conformance testing
11:47:57 [PhilA]
In our defence m'lud, we did discuss conformance more than 6 months ago, which is closely related.
11:48:01 [fadmaa]
bhyland1: discussing tools building should have taken place long time ago
11:48:23 [fadmaa]
cygri: I ask for testcases and not for tools
11:48:47 [fadmaa]
cygri: on the consumer side, testcases are example RDF graphs that uses the vocabularies and others that misuse
11:49:02 [fadmaa]
... on the producer side, SPARQL queries can be used
11:49:36 [fadmaa]
... potentially, the evaluator of some implementation might do some scripting to run queries or load the graphs
11:49:56 [fadmaa]
... that seams to be smaller burden than writing the testcase
11:50:05 [MakxDekkers]
sorry, have to leave now. good luck, talk to you again next week at our regular time.
11:50:18 [fadmaa]
11:50:24 [sandro]
PROPOSED: We'll build a test suite for each vocab. This will consist of example conforming and non-conforming uses of the vocab, and SPARQL queries which highlight conformance and non-conformance issues.
11:51:27 [fadmaa]
PhilA: we need to run the SPARQL queries against external data as well not only the ones we created
11:51:59 [DaveReynolds]
11:52:16 [bhyland1]
ack bhyland
11:52:33 [fadmaa]
cygri: asking for two people to confirm the conformance of some RDF data to a vocabulary is hard to be done
11:52:51 [fadmaa]
... there are cases where you need to be familiar with the reality the data describes
11:53:19 [fadmaa]
sandro: do you want to lower that and ask people to run the SPARQL queries on their data?
11:53:52 [fadmaa]
cygri: there is the general issue of how you test an implementation of a vocabulary
11:54:34 [fadmaa]
... if we have time and resources, how would we do it
11:54:48 [HadleyBeeman]
HadleyBeeman has joined #gld
11:54:52 [fadmaa]
... I'd like to discuss this regardless of time constraints we have
11:54:52 [sandro]
cygri: I'd to understand how to do it properly, and how well we can do it in the time we have.
11:54:54 [HadleyBeeman]
q+ Chair interrupt
11:55:21 [HadleyBeeman]
q- chair
11:55:25 [HadleyBeeman]
q- interrupt
11:55:27 [fadmaa]
cygri: then moving to discuss a possible compromise giving the WG time constraints
11:55:29 [HadleyBeeman]
q+ to chair interrupt
11:55:32 [bhyland1]
@PhilA, Agree, we did talk about conformance criteria and we also spoke about 2 ref implementations. However, the penny only dropped (for me) today about what it means for each term in the vocab to be used in at least two data sources by two people.
11:55:47 [PhilA]
rrsagent, draft minutes
11:55:47 [RRSAgent]
I have made the request to generate PhilA
11:55:51 [bhyland1]
… Further do we have two people, (selected by the WG for their expertise), who have time to inspect each data source and confirm they are used in conformance to the spec … all in the time we have remaining. I'm trying to be realistic.
11:56:15 [sandro]
test suite, something like: PositiveConsumerTest, NegativeConsumerTest (turtle files) PositiveProducerTest, NegativeConsumerTest (sparql files)
11:56:20 [DaveReynolds]
11:56:33 [HadleyBeeman]
lunch for one hour
12:51:53 [HadleyBeeman]
HadleyBeeman has joined #gld
12:56:29 [fadmaa]
fadmaa has joined #gld
13:00:15 [HadleyBeeman]
Shall we come back together?
13:00:51 [BartvanLeeuwen]
Zakim, call GLDMeetingRoom
13:00:51 [Zakim]
ok, BartvanLeeuwen; the call is being made
13:01:26 [sandro]
Zakim, who is on the call?
13:01:26 [Zakim]
On the phone I see no one
13:01:38 [BartvanLeeuwen]
zakim is still lost
13:04:10 [PhilA]
PhilA has joined #gld
13:04:17 [HadleyBeeman]
Topic: Discuss input for (successor of GLD WG)
13:05:14 [PhilA]
13:06:13 [DeirdreLee]
DeirdreLee has joined #gld
13:06:43 [PhilA]
13:07:08 [bhyland]
bhyland has joined #gld
13:08:13 [fadmaa]
PhilA: There is an indication that Rec track might not be the proper way to handle vocabularies.
13:08:18 [sandro]
phil: People who want something on come to and suggest it
13:09:17 [fadmaa]
... exists and has its own process of extention
13:09:35 [fadmaa]
... we have the plan to emulate this process
13:10:00 [fadmaa]
... the basic idea is to allow community groups to work on vocabulary. if the work proves successful
13:10:47 [fadmaa]
... the community group can submit a vocabulary for official W3C support
13:10:57 [fadmaa]
... if they can have consensus
13:12:21 [fadmaa]
PhilA: the upcoming group will be an advisory group not a decision group
13:13:05 [fadmaa]
... a community group can have a vocabulary under namespace
13:13:15 [fadmaa]
... consistency is a concern
13:13:27 [fadmaa]
... deprecation will be allowed but not deletion
13:14:04 [fadmaa]
... if there is no consensus, the reserved namespace can be used for other vocabularies
13:14:57 [fadmaa]
... I hope to see multilingualism support in the tool that manages
13:16:24 [sandro]
PhilA: We have the multilanguage expertise in the group; it's not hard.
13:16:32 [fadmaa]
sandro: in the version one of the tool, it will accept a Turtle file
13:16:57 [DeirdreLee]
Google hangout is on if anyone wants to join visually
13:17:29 [TallTed]
TallTed has joined #gld
13:18:57 [fadmaa]
HadleyBeeman: building a validator for vocabulary could be one thing that the vocabulary review committee works on
13:19:29 [fadmaa]
BartvanLeeuwen: a validator can be based on a set of SAPRQL queries
13:19:47 [fadmaa]
PhilA: sounds a better fit of the validation group
13:19:51 [gatemezi]
13:20:30 [fadmaa]
13:20:32 [fadmaa]
13:20:34 [HadleyBeeman]
13:21:22 [gatemezi]
fadmaa: Is there any limit of scope of the vocab in the new group?
13:21:49 [fadmaa]
fadmaa:wil the group have restriction on the scope of the vocabularies
13:21:56 [fadmaa]
PhilA: no
13:22:30 [fadmaa]
bhyland: external vocabularies deployed on non-production machines frequently cause problems
13:22:37 [fadmaa]
... when the machine is down
13:22:53 [fadmaa]
... having some redundancies to back-up this is very important
13:22:59 [gatemezi]
13:23:06 [fadmaa]
... ack me
13:23:10 [fadmaa]
ack me
13:24:04 [HadleyBeeman]
ack gate
13:24:05 [fadmaa]
gatemezi: LOV currently has a number of versions per vocabularies which are hosted on the OKFN
13:24:36 [fadmaa]
... you always can get a differnet version of the vocabulary if one is unavailable
13:25:43 [gatemezi]
ack gatemezin
13:26:05 [fadmaa]
bhyland: I suggest also talking to George Thomas who is behind
13:27:54 [fadmaa]
PhilA: there is an increasing need for validation as well.
13:28:15 [fadmaa]
PhilA: Makx is working on a profile of DCAT that will be recommended to the EC
13:28:44 [fadmaa]
PhilA: a workshop on validation will probably take place on validation
13:29:19 [bhyland]
FWIW, re: VRC - this is a very useful & needed service offering from the W3C and I see huge value/alignment with what is needed for persistence of vocabs & namespace and orgs (starting with US Government ) who should look to W3C for advice on 'how to' and possibly services beyond just strategic advice.
13:29:25 [fadmaa]
PhilA: in the London workshop next week, we want to break the silos between linked data, Json, and XML worlds
13:29:48 [fadmaa]
PhilA: the workshop is not specific to linked data but with open data in general
13:30:29 [fadmaa]
PhilA: Horizon 2020 will require funded projects to publish generated data in an open way
13:32:05 [fadmaa]
PhilA: the plan for the subsequent WG is not to be specific neither to linked data nor to government data
13:32:12 [fadmaa]
... but open data in general
13:32:29 [fadmaa]
... the workshop next week will provide us more input to help working on the charter
13:33:43 [fadmaa]
HadleyBeeman: you might want to consider, 1) versioing as it is not solved anywhere yet 2) quality which related to provenance but not only provenance. adding the ability to describe the data quality
13:34:53 [fadmaa]
... to point out the status of the data quality even when it is not perfect e.g. this data is 95% accurate but be aware that ...
13:35:03 [fadmaa]
... 3) the issue of discovery
13:35:17 [HadleyBeeman] slide 17
13:35:19 [sandro]
my ears perk up.... discovery of vocabs or of data sources?
13:36:16 [fadmaa]
... in the early days of the Web, people rely on maintaining lists of good quality web sites in an adhoc manner
13:36:32 [fadmaa]
... then search engines changes the scene
13:36:57 [fadmaa]
... it seems that having a list of datasets sounds similar to the inefficient method of early days of the Web
13:37:05 [PhilA]
See agenda for London
13:38:30 [fadmaa]
HadleyBeeman: Dan Brickley will be chairing a panel on discovery during the London workshop
13:38:43 [HadleyBeeman]
13:40:14 [fadmaa]
bhyland asks cygri about the process of populating the LOD cloud
13:40:46 [fadmaa]
cygri: it is a manual process and it doesn't address discovery
13:41:20 [fadmaa]
... only parts relevant to the diagram are recorded there
13:42:01 [bhyland]
agenda + for future GLD WG mtg - topic: discovery of LOD datasets starting with process Richard & Anja started in 2007
13:43:26 [bhyland]
NB: Open Data on the Web agenda, see
13:43:48 [fadmaa]
BartvanLeeuwen: I think there is much focus on the publication side of open and linked data but not on the consumption
13:44:02 [fadmaa]
... especially within the government itself
13:44:18 [fadmaa]
PhilA: we had a workshop last year on use of open data
13:45:08 [cygri]
13:45:21 [fadmaa]
... we are working on that within project proposals and planned workshops
13:45:48 [fadmaa]
PhilA: ... one of them is called "the business of open data"
13:46:36 [fadmaa]
BartvanLeeuwen: I am interested in consumption in general not necessarily making money out of data
13:46:38 [PhilA]
The business of open data
13:46:53 [fadmaa]
... I am aware of a number of duplicated offers within government
13:47:53 [fadmaa]
cygri: what would the expected output, in terms of W3C recommendations, of the group be?
13:48:22 [fadmaa]
... open data is a broad topic and it will be challenging to get consensus about what set of documents this group will produce
13:48:37 [BartvanLeeuwen]
13:49:19 [Zakim]
restarting Zakim-bot in 5 minutes to recover bridge state
13:49:36 [fadmaa]
... we have experienced this to a certain degree in this group (vocabs, directory, best practices).
13:50:22 [fadmaa]
cygri: with broad scope, It is challenging to sustain sufficient resources for all planned results
13:51:58 [fadmaa]
cygri: It is helpful to havie strands of works so that people can join the WG and work on a particular strand
13:52:11 [fadmaa]
13:53:19 [fadmaa]
PhilA: it is generally hard to sustain interest within working groups
13:53:30 [PhilA]
Possible topics for the new WG
13:53:32 [PhilA]
Open Data WG Charter notes
13:53:32 [PhilA]
Head for Persistent URI Recommendation
13:53:32 [PhilA]
- Input from GLD BPs
13:53:32 [PhilA]
- Input from UKGovLD
13:53:32 [PhilA]
- Other policies, such as Dutch
13:53:32 [PhilA]
- Identify an differences in sectors, e.g. government, enterprise, retail
13:53:32 [PhilA]
Guidance on what it means to conformance to a vocabulary.
13:53:33 [PhilA]
Description language for APIs? Ref DCAT discussion.
13:53:33 [PhilA]
Granularity vocabularies - frequency of updates, scale of map etc. ?
13:53:33 [PhilA]
13:53:33 [PhilA]
Publishing vocabs, clickable diagrams? schema -> SVG
13:53:33 [PhilA]
From UKGovCamp...
13:53:34 [PhilA]
Versioning, new version of old data
13:53:35 [PhilA]
Corrections to existing
13:53:36 [PhilA]
Quality related to provenance. Not the same as prov. It's not perfect but here it is.
13:53:37 [PhilA]
Discovery. We're still in the midset of repositories, lists. Indexing? DCAT?
13:53:38 [PhilA]
Barrier is not being able to find the data you need.
13:53:39 [PhilA]
Automatic generation of something like the LOD cloud.
13:54:50 [Zakim]
restarting Zakim-bot to recover bridge state
13:55:10 [fadmaa]
HadleyBeeman: maybe also APIs for usage of personal data
13:55:37 [fadmaa]
BartvanLeeuwen: I agree this is a hot topic
13:57:49 [sandro]
RRSAgent, pointer?
13:57:49 [RRSAgent]
13:58:23 [fadmaa]
cygri: who do you expect to join?
13:58:53 [fadmaa]
PhilA: when drafting the charter will keep in mind who will be interested to do something not only whether it is interesting or not
13:59:26 [Zakim]
Zakim has joined #gld
13:59:34 [HadleyBeeman]
zakim, this is gld
13:59:34 [Zakim]
ok, HadleyBeeman; that matches T&S_(GLD)3:00AM
13:59:50 [fadmaa]
bhyland: the workshop will help clarify possible interest and topics
14:01:42 [fadmaa]
TOPIC: CR exit criteria
14:02:31 [HadleyBeeman]
topic: CR exit criteria (part 2)
14:02:41 [HadleyBeeman]
From before lunch… cygri: there is the general issue of how you test an implementation of a vocabulary
14:03:02 [HadleyBeeman]
(and from before) cygri: I'd to understand how to do it properly, and how well we can do it in the time we have.
14:03:19 [HadleyBeeman]
5 min break
14:14:13 [HadleyBeeman]
Aaaand… we're back.
14:15:03 [PhilA]
scribe: PhilA
14:15:11 [PhilA]
scribeNick: PhilA
14:15:34 [PhilA]
Topic: CR Exit Criteria (conclusion)
14:15:54 [PhilA]
HadleyBeeman: We started to talk about how we can test our vocabs... Bart?
14:16:28 [PhilA]
BartvanLeeuwen: Richard said before lunch that we could create some queries that could be used by people to evaluate whether data has used the vocabulary properly
14:16:57 [PhilA]
... we thought over lunch that it shouldn't be too hard to create a Web service that would run queries against data
14:17:06 [DaveReynolds]
WG shouldn't set up test server, should be done by implementation reporters
14:17:12 [PhilA]
... with the proviso that the editors should provide the SPARQL queries
14:17:40 [PhilA]
BartvanLeeuwen: And then after the GLD, maybe a future WG could take this on to generalise it
14:18:04 [PhilA]
BartvanLeeuwen: So something like "this SPARQL query should list your organisations and the people within them"
14:18:39 [PhilA]
HadleyBeeman: Are the implementation reports not all WG members?
14:18:44 [PhilA]
DaveReynolds: Ideally not
14:19:16 [PhilA]
bhyland: Who would run the tests?
14:19:29 [PhilA]
DaveReynolds: For ORG we have several (UK, IT, EL etc.)
14:19:40 [HadleyBeeman]
draft table
14:20:00 [PhilA]
HadleyBeeman: For that point Dave that table above can be tailored to the specific example
14:20:21 [PhilA]
cygri: I don't see where automation is necessary
14:20:50 [PhilA]
DaveReynolds: You need a way of people saying that a given term has been tested
14:21:14 [PhilA]
cygri: With our vocabs we could probbaly do it manually as the vocabs are so small
14:21:39 [PhilA]
rrsagent, draft minutes
14:21:39 [RRSAgent]
I have made the request to generate PhilA
14:22:05 [PhilA]
DaveReynolds: One clarification - implementation reports get you from CR to PR, not into CR
14:22:16 [PhilA]
sandro: We're supposed to mention any examples when we go into CR
14:22:37 [PhilA]
HadleyBeeman: The list I have there comes from the process Doc
14:23:10 [HadleyBeeman]
"Are there any implementation requirements beyond the defaults of the Process Document? For instance, is the expectation to show two complete implementations (e.g., there are two software instances, each of which conforms) or to show that each feature is implemented twice in some piece of software?"
14:23:16 [PhilA]
HadleyBeeman: Having said that... hang on. We need to express any deviation from the standard ones in the process doc
14:23:27 [PhilA]
... we have to say what we will be showing, not what we can show now
14:23:27 [DaveReynolds]
To quote from Sandro's email "Is there a preliminary implementation report? The implementation report should be a detailed matrix showing which software implements each feature of the specification."
14:23:56 [DaveReynolds]
I'm saying we should list known implementations but not provide a preliminary implementation report at CR transition.
14:24:02 [PhilA]
cygri: If we want to create a test suite and base the exit criteria on that, then what we have to do now is to create the tests
14:24:19 [PhilA]
cygri: Then to get from CR to PR we need implementations that pass 9at least 2 or whatever we specify)
14:24:24 [PhilA]
14:24:47 [PhilA]
HadleyBeeman: So we need to lay out our exit criteria and our methodology
14:25:11 [PhilA]
cygri: We can also say now that we know about implementation A, B and C...
14:26:13 [sandro]
+1 DaveReynolds WG members don't have to do any testing.
14:26:16 [bhyland]
14:26:17 [PhilA]
DaveReynolds: This worries me. Other WGs don't do testing, they provide tools for implementors to do the tests
14:26:37 [PhilA]
HadleyBeeman: That makes sense to me. And removes some of the burden from us
14:26:55 [PhilA]
cygri: So if there is a broken implementation, that doesn't create a probelm from a process POV
14:27:07 [HadleyBeeman]
ack bhy
14:27:16 [PhilA]
cygri: if someone else has got it wrong then that shouldn't derail the problem
14:27:27 [PhilA]
14:27:46 [PhilA]
bhyland: What's the incentive to run the tests?
14:28:13 [PhilA]
cygri: Because the publishers want to promote the fact that they conform to the stnandard
14:28:35 [DaveReynolds]
14:28:49 [PhilA]
cygri: looking back to RDB2RDF test suite, people started using it as soon as we published it
14:30:02 [PhilA]
cygri: I don't know if it's governments that will do this, it could be others
14:30:04 [PhilA]
14:30:21 [PhilA]
bhyland: I think it will take a little out reach on the part of the WG
14:30:30 [PhilA]
bhyland: It's not trivial or insignificant
14:31:04 [DaveReynolds]
ack me
14:31:12 [PhilA]
sandro: Quick story - SPARQL was stuck on some bits that no one was interested to do. The I saw someone at a conference and a day later it was done...
14:31:51 [HadleyBeeman]
14:32:41 [HadleyBeeman]
bhyland: we should have an action item on this, one of us do the outreach of finding people to run implementation reports on their implementations for us
14:33:16 [HadleyBeeman]
phila: PwC will want to run tests against ORG and DCAT, for example
14:33:42 [HadleyBeeman]
14:33:45 [HadleyBeeman]
ack phila
14:33:50 [HadleyBeeman]
14:34:52 [HadleyBeeman]
cygri: the last bullet point is the most interesting one (re metrics to show that stuff has been implemented).
14:35:03 [PhilA2]
PhilA2 has joined #gld
14:35:12 [PhilA2]
scribe: PhilA2
14:35:30 [PhilA]
scribe: PhilA
14:35:37 [PhilA]
HadleyBeeman: So are we happy with the proposal?
14:35:48 [PhilA]
cygri: I'm not particularly happy with it
14:36:11 [PhilA]
cygri: I think we should make an attempt to create some test data
14:37:09 [PhilA]
rrsagent, draft minutes
14:37:09 [RRSAgent]
I have made the request to generate PhilA
14:38:04 [sandro]
(Earlier Test-Suite )PROPOSED: We'll build a test suite for each vocab. This will consist of example conforming and non-conforming uses of the vocab, and SPARQL queries which highlight conformance and non-conformance issues.
14:39:29 [PhilA]
DaveReynolds: My comment was about QB. It doesn't fit this pattern. I don't want people to run a server that people could run 1 million triples in QB when some queries are quadratic or more
14:39:51 [PhilA]
bhyland: So QBG is a unique animal. But Hadley has put together a template that seems to work well
14:40:21 [PhilA]
The burden of should not be on the WG members to do the test, but someone does need to facilitate recording the test result in the page
14:40:47 [PhilA]
bhyland: Someone needs to take an action item to facilitate the testing of each vocab
14:41:01 [PhilA]
cygri: Creating the test data is not trivial
14:41:15 [PhilA]
cygri: Asking people to run the tests aginst their data is mino
14:41:31 [PhilA]
cygri: By committing to creating a test suite we create a significant amount of work
14:41:46 [PhilA]
bhyland: Yes, that's why I think the bar is high and I'm abstainin
14:42:03 [PhilA]
bhyland: But Bart said he could build a service
14:42:16 [PhilA]
BartvanLeeuwen: Yep, I can host a service but the editors need to create the SPARQL queries
14:42:41 [PhilA]
bhyland: Not sure we're going to resolve this today, but it's clearly an issue
14:43:00 [PhilA]
sandro: I think we should try and decide
14:43:06 [PhilA]
bhyland: I agree there's an urgency
14:43:49 [DaveReynolds]
14:43:50 [PhilA]
sandro: I think there's consensus, if not enthusiasm, that we can do the human comparison thing
14:44:09 [PhilA]
cygri: I have to disagree here. The exit criteria about 2 humans checking doesn't work for me
14:44:36 [PhilA]
cygri: It's very subjective and it eitehr puts the burden on the Wg to do that or it puts the burden on the implementers
14:44:47 [PhilA]
... asking them to assert that their data conforms is too low a bar
14:44:55 [PhilA]
.... not happy
14:45:23 [PhilA]
sandro: I see the editors doing this but not with reading millions of triples
14:45:31 [PhilA]
14:45:41 [PhilA]
ack DaveReynolds
14:45:42 [cygri]
14:46:00 [PhilA]
DaveReynolds: I think I may be seconded with what Richard said.
14:46:11 [PhilA]
... who are those humans and what are their criteria
14:46:18 [PhilA]
... the danger is that the humans are me and Richard
14:46:42 [PhilA]
... I took a lot of time interpreting the conformance constraints for QB...
14:47:14 [PhilA]
... manbe there are some queries that we could write that would check for obvious problems
14:47:35 [PhilA]
... an unbounded amount of work is not good, no work isn't enough
14:48:19 [PhilA]
HadleyBeeman: So would it help you and Richard to keep in our WG advice that the inspection and confirmation process should be defined and documented case by case by the editors
14:48:39 [cygri]
14:48:53 [PhilA]
HadleyBeeman: I'm trying to put enough in the text to allow us to move forward without being too restrictive
14:49:34 [HadleyBeeman]
I'm suggesting we agree that the exact methodology to confirm each vocabulary is left to the editors to work out. They will find some combination of objective testing (e.g., SPARQL queries) and some subjective checking, as resources permit.
14:50:47 [PhilA]
PhilA: I've built a validator forJoinup's ADMS profile
14:50:58 [PhilA]
cygri: You can't give a complete test
14:51:09 [PhilA]
HadleyBeeman: But you can sample
14:51:15 [PhilA]
cygri: Yes
14:51:25 [DeirdreLee]
14:51:34 [PhilA]
cygri: I agree that we need a mix of using a test suite and a human
14:51:38 [PhilA]
14:52:43 [PhilA]
cygri: So regarding the manual inspection and confirming that they conform... it would be good to write a script that the tester could use to spot specific errors, and we should publish that as part of our documentation
14:53:03 [PhilA]
cygri: if we're doing a good job then we could perhaps ask implementers to do this themselves.
14:53:23 [DaveReynolds]
14:53:38 [PhilA]
... that would shift the burden. I'd rather write a checklist for others to use rather than check an unlimited number of data sets
14:53:58 [PhilA]
DeirdreLee: Would it be possible to use Gofran's tester that tests interoperability tester?
14:54:14 [HadleyBeeman]
14:54:19 [HadleyBeeman]
ack cygri
14:54:21 [HadleyBeeman]
ack deirdre
14:54:33 [DeirdreLee]
ack DeirdreLee
14:54:59 [PhilA]
fadmaa: I know it makes some assumptions about the communication that may not be valid
14:55:03 [PhilA]
ack DaveReynolds
14:55:38 [PhilA]
DaveReynolds: One of the things you want in the implementation report is a report on "we found it easy to use..." and we saw no use for ... etc
14:55:46 [PhilA]
... that's what's helpful
14:56:07 [PhilA]
DaveReynolds: I think we should include the means to add some descriptive text
14:56:22 [PhilA]
DaveReynolds: I think I'd find that more useful than eyeballs on triples
14:56:29 [HadleyBeeman]
14:56:34 [PhilA]
cygri: Checking whether terms have been used at all can be automated
14:57:30 [PhilA]
cygri: Commenting on the usability or usefulness... are there examples of others who have done
14:57:33 [DaveReynolds]
DaveReynolds: written feedback from implementers seems to me to be an important feature of implementation reports, what was hard to understand, what was useful, what didn't have value
14:57:44 [PhilA]
+1 to DaveReynolds
14:57:58 [PhilA]
DeirdreLee: Wouldn't that be the same as the comments from LC
14:58:00 [PhilA]
PhilA: Yes
14:58:29 [PhilA]
cygri: The difference about CR is that we want to know how it was actually used, not just thought about
14:58:34 [PhilA]
ack HadleyBeeman
14:58:44 [fadmaa]
fadmaa has joined #gld
14:58:45 [HadleyBeeman]
14:59:59 [PhilA]
cygri: I'm thinking of some sort of script that we want implementers to work through. That shifts the burden from the nebulous 2 humans to the implementers
15:00:12 [PhilA]
HadleyBeeman: proposes revised text...
15:00:16 [HadleyBeeman]
proposes: Each term in the vocab used in at least two data sources (ideally governments) and, using the test methodology, inspected each data source and
15:00:16 [HadleyBeeman]
confirmed they are used in conformance to the spec.
15:03:45 [HadleyBeeman]
15:04:49 [PhilA]
cygri: Org is the one where we're hitting these things first
15:05:03 [PhilA]
DaveReynolds: No... the current timetable has Org and QB at the same time
15:05:26 [PhilA]
cygri: I hear consensus on having SPARQL queries for each vocab
15:05:55 [PhilA]
cygri: These are about chekcing conformance on the publishing side. Do we need to worry about the consumer side as well
15:07:52 [DaveReynolds]
Data for unit tests are different from data from system tests
15:07:58 [DaveReynolds]
15:08:07 [HadleyBeeman]
PhilA: creating the SPARQL query alone will force the tester to create conformant and non-conformant data. That creates a reasonable test suite, useful for publishers and consumers.
15:08:21 [DaveReynolds]
15:08:28 [PhilA]
ack DaveReynolds
15:08:30 [HadleyBeeman]
PhilA: It'd be nice to get implementation reports from consuming applications, but not necessary for our exit criteria.
15:09:11 [PhilA]
DaveReynolds: I think it's a binary decision on whether we have to test consumers. We need to be clear whether we're including them or not
15:09:38 [PhilA]
DaveReynolds: It's not a trivial side efeect, or it certainly wasn't for QB
15:09:58 [PhilA]
cygri: That's right for QB, yes, but for Org it can be much simpler
15:10:11 [PhilA]
cygri: A well formed Cube is a much more complicated structure
15:10:39 [PhilA]
DaveReynolds: true - but I didn't need to create a test for every aspect of it
15:10:55 [PhilA]
cygri: No one suggests that the queries or the test data shoujld provide complete coverage
15:11:22 [PhilA]
... we prob want data in the reports that tell us whether each term has been sued but we don't need a test for it
15:11:29 [PhilA]
15:11:49 [PhilA]
cygri: It's not necessary to cover everything. It is useful to cover most of it
15:12:25 [PhilA]
HadleyBeeman: You were talking about the queries (the 80/20 thing). does that also apply to the human review?
15:13:00 [PhilA]
cygri: If you show me an RDF graph and asked is this conformant. I'd take a look and eventually come up with an answer but it may be hard to come up with an explanataion
15:13:15 [PhilA]
... is it using the right properties? is it extending it correctly?
15:13:27 [PhilA]
... I don't know how I'd quantify my thinking
15:13:47 [PhilA]
cygri: I'm suggesting a script or checklist
15:13:55 [PhilA]
... with binary answers
15:14:23 [PhilA]
... some questions might be like "are there any terms with the same semantics as an org term?"
15:14:45 [HadleyBeeman]
15:14:48 [PhilA]
... some inspectors will be more thorough?
15:15:27 [HadleyBeeman]
Can we look at this again?
15:18:15 [PhilA]
cygri: Will we have a test for every term?
15:18:26 [PhilA]
cygri: Each term is used by at least 2 data sources
15:18:41 [PhilA]
cygri: There are 2 data sources that pass manual inspection
15:19:23 [PhilA]
cygri: there are two sources that pass the test suite which consists of the queries and the human inspection
15:21:17 [DaveReynolds]
15:23:13 [PhilA]
ack DaveReynolds
15:23:45 [PhilA]
DaveReynolds: I am confused now
15:24:46 [HadleyBeeman]
PhilA: exit criteria: we will show that weher terms have been used, they have been used at least twice in line with the conformance criteria.
15:24:57 [HadleyBeeman]
… methodology is how we prove that
15:25:30 [sandro]
PhilA: Exit Criteria: (1) We will show that our terms have been used in conformance with the spec, and (2) at least two datasets have used each term
15:29:31 [sandro]
cygri: 1. We should show that every single term in the vocab has been used in at least two data sources. When doing this, we don't check if it's been used correctly.
15:30:46 [sandro]
cygri: 2. We will have a set of SPARQL queries, and each query comes with a description of the expected results. Like: this is a list of all the orgs mentioned in your data. We'll run these SPARQL queries and see if the results seem to match the description. We want to have two publishers that pass all of these.
15:31:15 [sandro]
cygri: Where it's really the Implementors running the queries, and reporting back to use about whether the results look correct.
15:32:24 [bhyland]
bhyland has joined #gld
15:32:38 [sandro]
cygri: 3. We in the WG will prepare a checklist or script saying which things need to be checked. Implementors will look at their data to see if it meets those criteria. For example: none of the non-org term has the same semantics as an org term. We'll say "look at your data and check for this".
15:33:01 [sandro]
HadleyBeeman: We need to define these test to go into CR.
15:33:16 [sandro]
cygri: Right, and we need two passing data sources.
15:33:52 [sandro]
cygri: I think these threee approaches, taken together, will give us an Implementation Report where it's not too much work on anybody, but we have confidence.
15:33:56 [sandro]
15:33:57 [sandro]
15:34:01 [DaveReynolds]
15:34:03 [fadmaa]
15:34:26 [HadleyBeeman]
PROPOSED: To adopt cygri's three points as above to be our exit criteria for the vocabularies
15:34:31 [DeirdreLee]
15:34:32 [sandro]
15:34:34 [BartvanLeeuwen]
15:34:36 [DaveReynolds]
15:34:36 [PhilA]
15:34:40 [fadmaa]
15:35:11 [HadleyBeeman]
PROPOSED: To adopt these as our exit criteria for vocabularies:
15:35:13 [HadleyBeeman]
1.  We should show that every single term in the vocab has been used in at least two data sources.   When doing this, we don't check if it's been used correctly.
15:35:40 [sandro]
RRSAgent, pointer?
15:35:40 [RRSAgent]
15:36:04 [HadleyBeeman]
HadleyBeeman has joined #gld
15:36:48 [HadleyBeeman]
PROPOSED: To adopt these three points (as described by cygri) as our exit criteria for vocabularies
15:37:04 [DaveReynolds]
15:37:07 [BartvanLeeuwen]
15:37:12 [fadmaa]
15:37:13 [PhilA]
15:37:14 [bhyland]
15:37:15 [sandro]
15:38:20 [DaveReynolds]
+0 (seems confused, for some vocabs the SPARQL queries will be yes/no, the boundary between inspection of SPARQL results and inspection of data seems awkward)
15:38:24 [PhilA]
rrsagent, draft minutes
15:38:24 [RRSAgent]
I have made the request to generate PhilA
15:38:52 [sandro]
15:39:26 [HadleyBeeman]
RESOLVED: To adopt these three points (as described by cygri) as our exit criteria for vocabularies
15:39:37 [DaveReynolds]
DaveReynolds: Overall seems to lead to an unconvincing result and a lot of work for the editors.
15:39:41 [sandro]
15:39:53 [DaveReynolds]
DaveReynolds: but current lack a better concrete proposal.
15:40:13 [DaveReynolds]
DaveReynolds: except that, as stated, QB will be different because of the existing IC tests and the need to test the normalization
15:40:28 [HadleyBeeman]
15:40:36 [cygri]
DaveReynolds, I agree that QB should be different. Was really thinking about ORG here
15:41:26 [PhilA]
HadleyBeeman: declares topic closed!
15:42:18 [PhilA]
HadleyBeeman: I will create CR pages for the other Rec Track documents. It would be good if editors cold start to populate those pages
15:42:30 [PhilA]
... if you haven't updated the timetable page for your docs, please do
15:42:34 [DaveReynolds]
Timetable will need to further extended given above resolution :(
15:42:35 [PhilA]
... and is that it?
15:43:32 [PhilA]
bhyland: Wraps up and thanks everyone for coming
15:43:43 [PhilA]
bhyland: Thanks DeirdreLee for organising the hosting
15:44:11 [DaveReynolds]
Bye all
15:44:15 [PhilA]
BartvanLeeuwen: I'd rather have 6 fires in a day - less exhausting than this
15:44:18 [Zakim]
15:44:27 [Zakim]
15:44:50 [Zakim]
15:45:03 [PhilA]
rrsagent, generate minutes
15:45:03 [RRSAgent]
I have made the request to generate PhilA
15:49:56 [DaveReynolds]
DaveReynolds has left #gld
15:52:57 [bhyland]
Ping Sandro, would you put today's minutes through commonscribe?
16:07:36 [HadleyBeeman]
HadleyBeeman has joined #gld
16:19:23 [cygri]
cygri has joined #gld
16:34:02 [PhilA]
zakim, bye
16:34:02 [Zakim]
leaving. As of this point the attendees were GLDMeetingRoom, Sandro, [IPcaller]
16:34:02 [Zakim]
Zakim has left #gld
16:34:07 [PhilA]
PhilA has left #gld
17:27:14 [cygri]
ACTION: Richard to write up justification for no-change approach to ISSUE-64 (Dataset vs dataset vs hasDataset)
17:27:14 [trackbot]
Created ACTION-119 - Write up justification for no-change approach to ISSUE-64 (Dataset vs dataset vs hasDataset) [on Richard Cyganiak - due 2013-04-19].
18:36:29 [HadleyBeeman]
HadleyBeeman has joined #gld