- From: Peter F. Patel-Schneider <pfpschneider@gmail.com>
- Date: Fri, 18 Jul 2014 01:21:38 -0700
- To: Jose Emilio Labra Gayo <jelabra@gmail.com>
- CC: Karen Coyle <kcoyle@kcoyle.net>, "public-rdf-shapes@w3.org" <public-rdf-shapes@w3.org>
On 07/18/2014 01:12 AM, Jose Emilio Labra Gayo wrote: > On Fri, Jul 18, 2014 at 9:49 AM, Peter F. Patel-Schneider > <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> wrote: > > So you are saying that ShEx is ambiguous as to whether open or closed > semantics is to be applied? That seems to be a problem with the ShEx > definition. > > > No, I am saying that there are two ways to implement Shape expressions...one > with the open shapes that allows remaining triples and one with closed shapes > which disallows them. That's what I said, I think. > The first implementation from Eric chose open shapes, while in my Shexcala > implementation I chose closed shapes. > > However, once I implemented it, I noticed that it was very easy to handle also > open shapes...and now, I have a flag in Shexcala so a user can select open vs > closed shapes when validating. > > In the future, my opinion would favour to have both open and closed shapes > with some syntax that allows one to indicate which one he prefers. For > example, I would propose: > > <A> { :a . } > > to be an open shape...it can have property :a with any value and other > remaining triples > > and > > <A> [ :a . ] > > to be a closed shape...it can have property :a and only property :a, and no > other remaining triples. > > In fact, this distinction between open and closed shapes is similar to regular > expressions (it is not a coincidence, given that ShEx is based on regular > expressions) where you can express that you can have an open regex like: "aa" > or a closed one like "^aa$" > > Best regards, Jose Labra > > > peter > > > > On 07/17/2014 02:05 PM, Jose Emilio Labra Gayo wrote: > > > On 7/16/14, 9:38 PM, Holger Knublauch wrote: > > Most people in my experience don't care about open world > semantics, but > of course nobody would admit that because it's against the > specs and > thousands of academic papers. > > > The cultural heritage community cares deeply about open world > semantics. > This community has a tradition of creating primarily > public-facing data > and, even in pre-Web eras, sharing that data widely. For the cultural > heritage community, the public, open web is the primary target > for its data. > > You confirm for me the impression that much of the discussion > here is in > the context of enterprise data systems. I will, however, do my > best to > keep the open world visible in these discussions. > > > I don't think that those 2 visions (open & closed world) are > incompatible. As > Kendall Clark noticed, constraint checking can also be combined with Open > World systems. One goal of Shape Expressions is to help in the > integration of > heterogeneous systems in an Open World. > > In fact, one of the first points in the discussion of ShEx was the > possibility > to declare open shapes instead of closed ones. The difference is that in a > declaration like: > > <PersonShape> { foaf:name xsd:string } > > and with the triples: > > :john foaf:name "John" . > > :mary foaf:name "Mary"; > foaf:mbox <mailto:mary@m.com <mailto:mary@m.com> > <mailto:mary@m.com <mailto:mary@m.com>>> . > > > a system with open shapes would assign both :john and :mary the shape > <PersonShape> because both have :foaf:name, while a system with closed > shapes > would only assign :john that shape, because :mary has an extra triple. > > Eric's implementation employs Open shapes, while Shexcala first employed > closed shapes and now admits both. > > In my opinion, closed shapes are good when you really need to ensure > that your > graph contains some triples and only those triples, while open shapes are > better in an Open World where you want to ensure that your graph has some > shape (if it has the triples declared in the shape) but it could also have > some remaining triples. > > So as a general remark, I really think the cultural heritage domain > can be a > very nice use case where the needs of integrating data from different > RDF data > portals appear. > > Best regards, Jose Labra > > > > kc > -- > Karen Coyle > kcoyle@kcoyle.net <mailto:kcoyle@kcoyle.net> <mailto:kcoyle@kcoyle.net > <mailto:kcoyle@kcoyle.net>> http://kcoyle.net > m: 1-510-435-8234 <tel:1-510-435-8234> <tel:1-510-435-8234 > <tel:1-510-435-8234>> > > skype: kcoylenet > > > > > -- > Saludos, Labra > > > > > -- > Saludos, Labra
Received on Friday, 18 July 2014 08:22:07 UTC