08:00:09 RRSAgent has joined #shapes 08:00:09 logging to http://www.w3.org/2015/09/09-shapes-irc 08:00:11 RRSAgent, make logs rdf-data-shapes 08:00:11 Zakim has joined #shapes 08:00:13 Zakim, this will be SHAPES 08:00:13 I do not see a conference matching that name scheduled within the next hour, trackbot 08:00:14 Meeting: RDF Data Shapes Working Group Teleconference 08:00:14 Date: 09 September 2015 08:00:32 present+ ericP, Arnaud, iovka, kcoyle, PhilA, hknublau, simonstey, pfps 08:00:46 chair: Arnaud 08:00:58 hsolbrig has joined #shapes 08:01:40 πρεσεντ+ διμιτρισ 08:01:44 present+ dimitris 08:01:59 I've another meeting @ 13:00 CEST but will return afterwards 08:02:06 ok 08:02:18 present+ dimitris 08:02:33 aryman has joined #shapes present+ aryman 08:03:07 Uh, guys... 08:03:25 the phone number above appears to be a phone sex line. 08:04:04 that's just Arnaud talking sexy 08:04:21 Never mind. 08:04:26 which one? - not that I'm saying that you are wrong, but I'm listening to the people in Lille :-) 08:04:58 Arnaud_ has joined #shapes 08:05:37 scribenick: dimitris 08:05:39 didn't that bleed into the IRC? I see something very strange from Dimitris 08:06:48 that's Greek to me topic: Compact syntax 08:08:12 arnaud: we decided to work on a compact syntax along the lines of shex 08:09:12 .. Eric has some material he wants to present 08:09:20 q? 08:09:58 http://www.w3.org/2015/Talks/0909-shex-egp/#%281%29 08:11:18 ericP: (presenting the slides) 08:12:42 Labra has joined #shapes present+ labra 08:13:55 As far as I know, ShEx repetition is not a part of SHACL 08:14:32 q+ 08:14:59 This appears to be strangely missing the possibility of following property links 08:15:29 ack pfps 08:16:19 q+ 08:16:22 pfps: this page (slide 3) is missing the idea of following property links 08:16:43 actually that is hidden as part of "triple constraint" 08:16:58 ... (ok found it) 08:17:02 Can we settle on an agreed term for a collections of shapes?. i.e. schema, shapes graphs, ... 08:20:54 The basic algebra of ShEx is very different from SHACL - conjunction vs addition 08:23:40 EXTRA doesn't add any power, as far as I know 08:23:41 ericP: (answering aruman on slide 3) extra allows other values 08:24:40 q? 08:24:58 ack aryman 08:24:59 doesn't SHACL handle IRI stems using patterns? 08:26:06 doesn't SHACL handle inclusion via AndConstraint? 08:26:20 ... semantic actions can do validation among other things 08:26:26 SHACL has EXTRA on by default 08:26:28 q+ 08:27:20 as far as I know the only significant differences between SHACL and ShEx is the combination algebra and semantic actions 08:27:47 yes, ShEx partitions, and this makes ShEx processing hard 08:28:22 aryman: you have to compute all the partitions of each triples? 08:29:05 ericP: yes, when compute multiple occurences 08:29:38 iovka: we have a smart algorithm, worst case is np-complete 08:30:34 q+ to ask if Iovka's algorithm should be in the spec? 08:30:47 ack pfps 08:30:47 aryman: this feature is different from shacl 08:31:37 pfps: shacl supports IRI stems from patterns and inclusion from and constraints 08:31:52 q- 08:32:11 Arnaud has joined #shapes 08:32:23 q? 08:35:56 pfps: you can achieve this using a union 08:36:45 ericp: the question is to figure out how to compile shex into shacl 08:37:04 pfps: shex doesn't have conjunctions 08:37:29 pfps: SHACL has both a universal and a counted construct 08:38:38 pfps:you 2-4 shoes that belongs to class A and 2-4 shoes belong to class B 08:38:57 .. in shex algebra you would count it once 08:39:05 pfps: as far as I can see the biggest difference between SHACL and ShEx is the combination algebra 08:39:53 iovka: if you want the same and it is additive then you would have to add negation 08:40:11 pfps: yes you would have to exclude the middle 08:40:58 iovka: I don't know what happens in you have unbound cardinality 08:41:04 pfps: in SHACL, values satisfiy different constraints independently; in ShEx, values can only satisfy one of a set of constraints 08:43:28 ericP: inclusion is like substituting, more like a macro 08:43:44 aryman: is there nesting in shex? 08:44:02 again, SHACL has this via AndConstraint, 08:44:14 ericP: you can do nesting with AND and OR in the same way with shacl 08:46:07 it's not an "AND" it's a "PLUS" 08:46:27 aryman: is there an implicit and and if you want an or you would explicit state it? 08:46:35 ericP:yes 08:48:04 aryman: your meaning of AND is different than SHACL 08:48:54 Arnaud has joined #shapes 08:48:55 q? 08:48:58 ericP: on multi occurrence it is obvious 08:51:18 q? 08:54:30 Arnaud_ has joined #shapes 08:55:27 q+ to ask whether slide 9 is shex-to-json or shex-to-SHACL 08:55:41 q- 09:00:17 q? 09:00:18 ericP: (discussing with pfps about ways to translate between shacl and shex) 09:01:29 ... it is easier to map from shex to shacl (rdf) 09:02:25 Is there a connection from here to a user-friendly syntax for SHACL? 09:04:07 arnaud: we have two different animals and we need a way to combine shex with shacl and how we deal with differences 09:04:14 q+ 09:04:52 q+ 09:05:26 ack kcoyle 09:06:05 kcoyle: is shex a user-friendly language? we didn't see much syntax 09:06:12 q+ 09:06:37 ack pfps 09:06:46 ericP: we shown the syntax quite a few times in the past 09:07:23 pfps: shex syntax is fairly compact, easy to write & read, has a few peculiarities 09:07:44 +q 09:08:31 ... are the shex people in favor to remove conjuction? 09:09:01 to ask whether the shacl people are ready to turn conjunction into grouping ? 09:09:40 ericP: conjuction is the major issue 09:11:12 ... conjunction breaks user intuition. if shacl has features that are better for users we can go in that direction 09:11:31 s/conjuction/conjunction/ 09:12:13 ack aryman 09:13:14 aryman: I am with peter. I find it odd that people in this WG that develop their own tools. Let's see the differences and create a unified language 09:14:16 ack iovka 09:14:44 iovka: what would it take for shacl people to translate conjunction in to grouping? 09:14:49 q+ 09:15:13 ... we should avoid having the same syntax for different languages 09:15:56 ack PhilA 09:16:29 q+ 09:16:37 q- 09:16:43 q+ 09:16:47 PhilA: good to see consensus, the will be one W3C standard and we need to converge 09:16:49 ack pfps 09:18:17 pfps: I don't see compromise reg. conjunction. Ideally we could add both approaches but not easy to implement 09:19:39 q+ 09:19:47 arnaud: we can discuss with examples and test cases to see both approaches and decide later 09:20:51 ack pfps 09:21:48 pfps: in the RDF(S) view point conjunction in the only way to go 09:22:40 q+ 09:22:48 ... shex is a difficult language from a computation perspective 09:23:07 +q 09:23:51 ... according to iovka it is np-complete in worse case or maybe more 09:24:41 according to iovka it is in NP, perdiod. not more 09:24:54 ... if we go for a union someone has to make the argument if the computation and implementation cost worth it 09:25:06 ack aryman 09:26:00 aryman: shacl is consistent with oslc 09:26:45 ... we have to understand the reality and see use cases 09:27:50 ... regarding complexity I trust iovka on optimizations 09:28:22 ack iovka 09:29:12 iovka: people are working on np-complete every day. 09:30:11 ... academics found complexity problems in SPARQL property paths and OPTIONAL 09:30:34 ... in practice worst-case complexity doesn't happen 09:30:46 q+ 09:31:34 q+ 09:31:57 arnaud: we can try to converge point by point 09:32:24 ack Labra 09:32:34 labra: the differences are not that big 09:32:57 ... only in multi occurrence 09:33:24 ack PhilA 09:33:43 PhilA: it's a year since you started 5 months ago 09:34:18 ... you published the UCR and I hear there are new use cases 09:34:33 ericP: the use case document does not capture this distinction 09:34:58 ... we have these in the wiki 09:35:12 PhilA: should be included in the UCR document 09:36:14 ... you have 1 year left 09:36:49 Correction, you have nearly 2 years left (to June 2017) 09:37:40 arnaud: the ball is in shex people, the wg is open 09:38:54 ericP: if we can include multioccurrence it's fine, otherwise there is nothing else shex can give to the WG 09:39:59 ... multioccurrence is a very common pattern in shex 09:41:11 ... only lately we understood that there would be a single semantics for multi occurrence 09:41:16 which decision made two weeks ago? 09:42:49 @pfps qcrs 09:42:51 q+ 09:43:07 aah - four weeks ago 09:43:15 arnuad: I am concerned with take it or leave it approach. The path forward is clear. you need to make an effort to convince the group that your approach is better 09:44:07 I believe the main difference is that SHACL does not include the GroupShape operator defined in http://w3c.github.io/data-shapes/semantics/ 09:45:05 it appears to me that the current SHACL semantics handles S47 09:45:10 ack aryman 09:45:44 ... which is the use case underlying the multioccurence issue - ISSUE-53 09:46:03 aryman: GroupShape is what is missing from shacl. If we add that to shacl would it satisfy you 09:46:17 The decision was: http://www.w3.org/2015/08/13-shapes-minutes.html#resolution04 09:46:35 ericP: that's true 09:46:37 The open issue was closed at that meeting 09:47:11 iovka: adding groupShape might be possible but we need to understand the semantics of the rest of shacl 09:47:24 And the issue was about multi-occurrance of the same predicates 09:47:48 the underlying operations in ShEx are different from those in SHACL 09:48:11 q+ 09:48:25 q+ 09:48:47 ack kcoyle 09:49:27 kcoyle: does the first working draft depend on this? 09:49:30 ack PhilA 09:50:08 q+ 09:50:54 PhilA: it's not good to hear from any WG member "either we do this or we walk away". We all need each other and have to do compromises 09:50:55 ack pfps 09:52:09 pfps: sorry if I sound bitter but the WG has been working without the shex people for some time now and we are not less viable 09:52:39 ... shacl is a result of quite a few compromises already 09:57:03 ... one solution would be to combine rdfs with recognition but it requires an owl reasoner and I don't think anyone would go with this approach 09:58:32 I perfect way to proceed here is to show that S47 is not currently handled in SHACL 09:59:38 q+ 10:00:28 ack pfps 10:00:56 RRSAgent, draft minutes 10:00:56 I have made the request to generate http://www.w3.org/2015/09/09-shapes-minutes.html PhilA 10:01:21 pfps: the basic design in shacl had one change with the qualified cardinality 10:01:44 ... in the past months 10:02:51 ericP: do I have to prove that the behaviour in slide 4 would act different in shacl? 10:03:24 pfps: you need to show that there is a use case that cannot handled by shacl 10:04:35 ... S47 is a good example 10:04:39 -> https://www.w3.org/2014/data-shapes/wiki/User_Stories#S47:_Clinical_data_constraints S47 10:08:01 http://www.w3.org/2015/Talks/0909-shex-egp/#%284%29 10:09:00 ericP: (discussing about slide 4 with Peter and Arthur) 10:15:23 I'll join again later this afternoon 10:57:15 simonstey has joined #shapes 11:10:04 kcoyle has joined #shapes 11:12:30 Arnaud has joined #shapes 11:12:36 PhilA has joined #shapes 11:12:50 we're back on, you guys call in 11:13:34 aryman has joined #shapes 11:14:21 present+ hknublau 11:15:55 scribenick: hsolbrig 11:16:24 arnaud: we need to know who is in charge and who we can depend on to lead the test suite and what needs to be done 11:16:56 Labra has joined #shapes 11:17:00 https://github.com/shexSpec/shexTest 11:17:25 PhilA_ has joined #shapes 11:17:52 ericP: the goal of the ShEx test suite is to test a given node in a given graph on a given shape in a given schema 11:18:25 present+ pfps 11:18:30 ericP: Fail or pass, but if pass we've defined a preferred solution so we can move the "semantic actions" out of band 11:18:50 ... we expect this group will be interested in just pass or fail 11:19:19 ... covers ShEx language features first, but will be happy to do the non-ShEx SHACL stuff as well 11:20:06 ... one issue is to what degree we can standardize errors. If we treat it as yes, here's a proof and no, here's an error, we need consistent error reporting 11:21:19 Arnaud: XML Schema stayed away from trying to dictate how an implementation functions. Have an open issue on results and are having second thoughts on whether we should go there at all... 11:22:00 Arnaud: If spec says "This error returns this information", we need to test that, but maybe simpe T/F is good enough for the test suite 11:22:19 iovka has joined #shapes 11:22:58 q+ 11:23:13 q+ 11:23:19 ericP: We haven't done anything in ShEx yet. True is give a proof, false is just false 11:23:31 ack pfps 11:23:53 pfps: It is not "errors" - it is violation reports. 11:24:30 ... it is very important to get back information on the violations. 11:24:53 ... test suite needs to say what violations are reported. 11:26:05 q+ 11:26:28 ... example, SHACL says "Hey, check whether everyone has a name", an answer of "No" isn't useful. 11:27:08 ericp: Does this work when you get into complicated shapes? 11:27:59 Arnaud: If I have a graph with 2 violations, do both need to be reported? 11:28:10 pfps: Editor's draft says what happens there. 11:28:31 https://github.com/w3c/data-shapes/blob/gh-pages/data-shapes-test-suite/tests/features/core/manifest.ttl 11:28:50 q- 11:28:55 Arnaud: Some implementation may keep on going, some may report first error only 11:28:56 +1 to Arnaud 11:29:03 ack hknublau 11:29:09 q- PhilA 11:29:37 hknublau: We've had resolutions on that topic and want the ability to specify details including invalid value and details. Has been decided. 11:30:11 ... look at test cases in link above, each can have more mf result objects, data that comes back from validation 11:31:19 I would view SHACL engines that don't produce violation results as broken 11:31:46 ericP: Answer seems to be there. Engines that are just interested in T/F can just detect and don't need to emulate details. 11:32:31 hknublau: I have a solution, but we should really have a different pair of eyes build the test cases. 11:32:44 q+ 11:33:14 ericP: I can work on the test suite and maintain it. 11:33:19 ack kcoyle 11:34:27 kcoyle: Issue 51, types of validation results, only stated type of validation is open and people could develop their own. Don't know about details 11:35:06 Arnaud: Details are part of issue 51 discussion 11:36:08 kcoyle: Would love to provide tests, but don't have SHACL writers, how can we provide 11:36:41 (BTW the test results are an example where a detailed data model of constraints (schema) may be needed - each violation should point back to the constraint instance, e.g. the object holding the minCount property). 11:36:47 ericP: We have ShEx to SHACL translator. 11:37:10 s/ericP/Labra/ 11:37:10 s/ericP: We have ShEx to SHACL translator./labra: We have ShEx to SHACL translator./ 11:37:19 Labra: In general, translation works. 11:37:57 Arnaud: We need to know what is needed to contribute a test 11:38:39 Labra: Some new things in ShEx I don't know how to translate to SHACL... 11:38:50 ... don't what the methodology will be to add new tests. 11:38:55 A tool to convert between syntaxes: http://rdfshape.herokuapp.com/converter/schema 11:39:12 holger has a lot of tests for his SHACL API 11:40:13 hknublau: Link to manifest file is all we have right now... the rest haven't been integrated. 11:40:50 hknublau: ... new tests would be new folder w/ manifest file w/ results and instructions along with tests and data 11:41:22 ... data is in one folder up from manifest 11:41:30 @jose Bad request For request 'GET /converter/schema' [Missing parameter: schema] 11:42:23 ... I've put SHACL and data in same file, but they don't have to be. I wish results would be in same file but... 11:42:49 http://rdfshape.herokuapp.com/ 11:43:14 That's the right link 11:43:37 ... there is a meta-manifest file a couple of folders up that you need to add manifest to 11:45:29 thx 11:45:32 Example of a conversion from compact syntax to SHACL (Turtle): http://goo.gl/JnAJiT 11:45:41 ericP: Manifest format comes from multiple groups. Usually run with manifest file, meta-manifest file, instance data and some representation of results. Tests generate "ERL" (eval and reporting language) reports 11:46:01 ... pushes off the task of result interpretation 11:46:26 Arnaud: We have a framework, Eric will contribute tests. 11:46:34 q+ 11:46:53 ericP: Tests are amethodical exploration of the language. 11:47:20 Arnaud: How much do you think you really cover in the language? 11:48:25 ericP: I'm about 10% of the way through. Have parser tests for ShEx and then several different tests for inputs. Meant to be pretty rigorous.. 11:48:38 ... also have tests to cover different logical permutations 11:48:42 ack hknublau 11:49:13 hknublau: If the process just becomes a dump from ShEx, not helpful. Should be custom designed for SHACL, not ShEx 11:50:00 ... someone needs to try to come up with test cases from the SHACL point of view. I don't want to have to guess whether this test is supposed to work or not. 11:50:28 Arnaud: I'm not going to turn down offers. Are there other volunteers? 11:50:54 I expect to be contributing tests, but probably not systematically 11:51:23 ... we should start with ShEx and find areas that need to be beefed up for SHACL 11:51:55 ... we need to resolved issues as tests. 11:52:24 ericp: In SPARQL, an issue was presented with two mutually exclusive tests -- only one of which would pass depending on how the issue was resolved. 11:53:15 Arnaud: Will you contribute as you go or will it be incremental? 11:54:38 ericP: naming conventions, etc. require periodic refactoring ... easier to do in one place ... don't want to have vestigial crap lying around from renaming. 11:55:47 Arnaud: as ShEx will merge w/ SHACL, can't we just have one repository? 11:58:02 ericP: In the DAWG we had tests that said "Unextended SPARQL did this" but we also had tests that could be annotated as being extended by specific features. We could do that here 11:58:25 q+ 11:58:26 ... e.g. EXTRA being ShEx only 11:59:06 ... testing could be modal. 11:59:23 ack PhilA 12:00:26 PhilA: Part of a test case is use cases, use cases should be part of test suite. 12:01:53 ... RDF data that you are checking could be submitted w/o necessarily writing ShEx. Data could or be provided that did and didn't pass use cases. 12:04:29 pfps: Lets get Karen to dump some examples and the rest of us can circle around it to build SHACL tests. 12:05:49 Arnaud: Can Karen create a sub-directory under tests directory and start adding data? 12:06:47 Kcoyle: Will create identified directory (e.g. dcmi) underneath -- so people can track source 12:08:16 Arnaud: How do people feel about eric using a repo w/ stuff that belongs to ShEx only? 12:08:54 hknublau: I'm skeptical. Will lead to a dump of files that don't work and will require a lot of energy. Wouldn't count on this happening or working 12:09:31 hknublau: Would recommend that actual test cases only use TTL. ShEx will just be parser tests 12:09:47 as long as there is a clear distinction saying what is what how does it harm things if there is some other stuff there 12:10:00 hknublau: We shouldn't force anyone to use compact syntax. 12:10:25 Arnaud: Will keep repo separate for now. 12:11:40 hknublau: By the way, I'm withdrawing objection to yesterday's resolution on the specification including SPARQL definitions. 12:13:05 topic: ISSUE-44 12:13:05 issue-44 -- How to express dependencies between graphs -- open 12:13:05 http://www.w3.org/2014/data-shapes/track/issues/44 12:14:21 hknublau: Suppose you want to validate something against a controlled list of terms (e.g. country codes). 3 graphs -- reference data, definition of macro and data graph itself. How can a tool that just gets a data graph 12:14:36 ... get instructions on how to get what it is supposed to use? 12:15:36 ... I've suggested a mechanism similar to OWL:Import w/ local mechanisms. 12:16:10 q+ 12:16:35 ack pfps 12:17:19 pfps: RDF should have resolved this but they didn't. 12:18:16 .... does SHACL need to again, take out this "RDF garbage" or should we just say "not our problem -- marshalling is done elsewhere" 12:18:45 ... another option would be to supply information in invocation 12:19:17 ... ya option would be to just use owl:imports 12:19:53 ... problem is if data or control IS OWL, what should owl do? 12:20:46 q+ 12:20:59 ... no good way forward. Only a least bad way, which is to "be an ostrich" and say not our problem. 12:21:04 ack hknublau 12:22:20 one problem with an imports directive is how to handle redirection 12:22:38 q+ 12:22:40 hknublau: Two different types of imports, can't just rely on owl:imports 12:22:46 q+ 12:22:53 scribenick: ericP 12:23:03 Arnaud: why do we need to have two different imports? 12:23:12 q- 12:23:18 ... why is it important to distinguish the import of data or schema? 12:23:28 hknublau: the opperation requires two graphs 12:23:53 ... the transitive closure of the imports graph would create the input graph 12:23:59 ack aryman 12:24:05 ... we don't want the validation to always include the shapes graph 12:24:31 aryman: I thought we already had a property that linked a graph to a shape... 12:24:32 aryman: why do we need another property to connect data to shapes? 12:24:42 ... in a linked data world, you can just follow the link 12:25:10 ... i think that dictating how you build up the data graph is outside the WG 12:25:23 hknublau: of course it would be nicer to have the import in the rdf namespaces 12:25:43 ... in many cases, you can't dereference. 12:26:07 aryman: the assumption in XML was that your document could point to a schema 12:26:28 ... you didn't need to GET that graph. 12:26:50 ... there's a standard for XML catalogs supported by e.g. Protege 12:27:05 ... can we standardize schema, shapes graph 12:27:35 hknublau: pfps uses "control graph" but there are other controls 12:27:52 aryman: so why another imports? 12:28:24 hknublau: so in TBC, a user starts an empty file and they need to link to the shapes graph 12:28:50 aryman: don't we have a property that links a graph to a shapes graph? 12:29:17 ... these URIs should be dereferencable. if they're not, we need an e.g. catalog mechanism 12:30:03 Arnaud: when we talk about inferencing, we say "it's outside shacl. is there's inferencing to be done, do it before. likewise e.g. HTTP GET" 12:30:25 ... so we can say "we don't go there; it must be done before" 12:30:42 hknublau: we have a hint property called sh:entailment 12:31:00 ... nobody's forced to use them. 12:31:11 q+ 12:31:36 ... i feel we're not looking at my needs 12:32:01 Arnaud: by the same token, if no one else supports it, what's the good of calling it a standard? 12:32:10 ack pfps 12:32:53 pfps: do we have a use case for this? 12:33:38 q+ 12:33:49 ack Dimitris 12:34:12 Dimitris: i find this useful but i would limit it to certain graphs 12:34:30 ... so if i have a library of definitions, it would be nice to include other shapes 12:34:46 ... but i wouldn't extend that to the instance graph 12:35:22 ... i think the data import is out of scope 12:35:32 hknublau: so use owl:imports? 12:36:01 I also agree that being able to use external information in shapes graphs is useful - I think that there is a use case on this somewhere 12:36:10 Arnaud: apart from the fact that we're not using anything from OWL, what would happen to e.g. composer? 12:36:52 hknublau: TBC respects owl:imports. at a minimum, i'd need the shapes import 12:36:54 q+ 12:37:35 Arnaud: there seems to be less resistance to the shapes import? 12:37:36 q+ 12:37:39 ack aryman 12:37:47 hknublau: then how do we distinguish a shapes graph? 12:37:48 https://www.w3.org/2014/data-shapes/wiki/Requirements#Including_Named_Graphs_for_Query_Evaluation 12:38:06 q+ 12:38:18 q+ 12:38:21 aryman: if we adopt "library", can we call it something other than "library" and define it beying a grab of triples? 12:38:34 hknublau: spec out of data on his 12:38:45 ... [ details on RDFS for this ] 12:39:26 ... if TBC encounters a link to a link, it dereferences it "somehow" 12:39:44 aryman: you also need the library include in the shapes graph? 12:39:45 ack pfps 12:39:52 hknublau: use for e.g. country code example 12:40:36 pfps: there was a dc (or cataloging)-related one for pointing to a slew of valid stuff 12:40:58 ... partly to reuse and partly because they were painful to include 12:41:13 ... i would not be opposed to a mechanism to build up the shapes graph 12:41:35 kcoyle: since pfps ref'd cc, most of our stuff is too big to actually be included. 12:41:56 ... we deference one URI at a time 12:42:18 ... some of it isn't easily includable 12:42:35 ... we pull in pieces at a time, not the whole file 12:42:58 q+ 12:43:03 pfps: you want the validation process to work in an impoverished env 12:43:15 kcoyle: we live in an impoverished env 12:43:35 ... our programmars are assuming they'll have to do the imports manually 12:43:52 ... we don't have a mechanism to know what we have to derefernce 12:44:06 Karen talks about http://www.w3.org/2014/data-shapes/track/issues/80 12:44:34 pfps: so taking LoC, we have a shape that says your LoC catalog number has to be in the reference 12:44:42 kcoyle: order 300M 12:45:23 pfps: so the use case is going to be that "the value exist in the value set", and thus painful to include 12:46:05