RE: Referencing FOAF

Biplav,

Thanks for this.  I agree with you.  I hope an effort to look at a catalog will be forthcoming.

Dan

From: Biplav Srivastava [mailto:sbiplav@in.ibm.com]
Sent: Tuesday, December 18, 2012 10:28 PM
To: Makx Dekkers
Cc: Gillman, Daniel - BLS; 'Phil Archer'; 'Public GLD WG'; 'Sandro Hawke'
Subject: RE: Referencing FOAF


Makx, Dan, et al.

My two cents because I see this issue cropping up repeatedly while linking data in new domains and geographies.

 *   Linking data to vocabularies is in itself a big effort with existing tools. So, people rarely make the difference between predicate or value vocabularies. In fact, people link to what they consider as the best fit at that time and from trustable sources.
 *   So, a mechanism to register/ version vocabularies will go a long way in promoting reuse. It is consistent with another area where the community has experience -  the approach of registering services (for reusing web services).
 *   If newer versions of vocabularies come in future, it is the choice of the vocab developers to support backward compatibility or the consumers to move to the newer version (with any support the platform provides for discovering new versions).
 *   It is expected that if there are choices of existing vocabularies that are well maintained (choice [2] in Makx note) or of unknown origin (choice [1]), unconstrained, rational consumers will first choose [2], and only if that does not work out, choose [1] or create their own. One would choose [3] for purely tactical reasons.
Regards,
--Biplav




From:

"Makx Dekkers" <mail@makxdekkers.com>

To:

"'Gillman, Daniel - BLS'" <Gillman.Daniel@bls.gov>, "'Public GLD WG'" <public-gld-wg@w3.org>

Cc:

"'Sandro Hawke'" <sandro@w3.org>, "'Phil Archer'" <phil@philarcher.org>

Date:

12/19/2012 02:00 AM

Subject:

RE: Referencing FOAF




________________________________



Thanks, Dan.

You make very valid points, but I think those relate mostly to what I would call “value vocabularies”, things you use in the Object position in a triple, for example sets of concepts like the Library of Congress Subject Headings, EuroVoc terms, Geonames etc.

My message was more narrowly focused on predicate vocabularies, like DCMI terms, FOAF, DCAT, ADMS and was triggered by the discussion in the call on 6 December on whether a W3C specification could rely on an external predicate vocabulary like FOAF that is not managed according to W3C rules and procedures.

But for value vocabularies, I agree that there is no sensible way to define “good” in the sense of “fit for purpose”. However, there could still be scope for defining “good” in terms of “well-defined and well-maintained” but maybe that’s not a subject that is in scope for this group.

Makx.



From: Gillman, Daniel - BLS [mailto:Gillman.Daniel@bls.gov]
Sent: Tuesday, December 18, 2012 8:38 PM
To: Makx Dekkers; Public GLD WG
Cc: Sandro Hawke; Phil Archer
Subject: RE: Referencing FOAF

Makx et al,

This issue is one that is near and dear to my heart, so I want to weigh in.  I think the only option that makes sense is #1 for the following reasons:

1)      Trying to establish some authority or process for declaring “good” vocabularies is a non-starter.  One person’s  “good” won’t be everyone’s, and there are always going to be quirky local reasons for choosing one vocabulary over another.  An example from my business (national statistical office)  is industrial classification systems.  These are vocabularies in the GLD sense, and there are many of them, including ISIC (International System for Industrial Classification).  A determination that ISIC should be used for all applications where an industrial classification is required would ignore the special needs of regional and national systems geared towards the specifics of those economies.  Same with declaring any one of the regional systems “good”.  So declaring “good” is problematic.
2)      When a vocabulary is referenced or used, mostly it is with the concepts in that vocabulary defined at the time of use.  But, for any popular vocabulary, over time changes will be made.  Concepts are redefined, split, or combined.  New concepts are incorporated.  But, none of those changes should invalidate the use that was made before the changes.
3)      It is rare that no existing vocabularies will suffice for the needs for some problem, meaning creating one’s own vocabulary is rarely needed

If we were to recommend the creation of a catalog of vocabularies, this I believe would solve many of the problems described below.  First, any time-stamped vocabulary could be registered, rendering the problem of updates moot.  Second, a catalog will support discovery, and the problem of knowing which vocabulary to use and whether there are any relevant possibilities will be much easier to solve.  The use of a catalog will substantially reduce or eliminate the risk that vocabularies will be over-written and invalidate data.  Third, the issue of conceptual overlap among  similar concepts across vocabularies can be recorded, enabling developers to map across vocabularies and provide deeper semantic links across data.

I submitted 3 documents early in the GLD process that address these issues.  I updated the available versions, and they are located at
1)      http://www.w3.org/2011/gld/wiki/File:Open_Government_Vocabualries_-_Registration_Procedure_v03.doc

2)      http://www.w3.org/2011/gld/wiki/File:Open_Government_Vocabualries_-_Content_Model_v02.doc

3)      http://www.w3.org/2011/gld/wiki/File:Open_Government_Vocabualries_-_Registration_Model_v04.doc


The first and third documents are especially relevant for this discussion.

Unfortunately, I will not be able to join the call on Thursday, but I can join during January (3rd or 10th).

Yours,
Dan



From: Makx Dekkers [mailto:mail@makxdekkers.com]
Sent: Tuesday, December 18, 2012 11:03 AM
To: Public GLD WG
Cc: Sandro Hawke; Phil Archer
Subject: Referencing FOAF

Dear all,

There was a short discussion in the GLD call two weeks ago (http://www.w3.org/2011/gld/meeting/2012-12-06) about referring to other people’s vocabularies. Here are some of my thoughts on this issue.

In earlier work that I was involved in, this issue has come up a couple of times. From those discussions, I remember three potential strategies:

1.       Re-use existing vocabularies irrespective of how they were developed and who is responsible for them. This has the advantage of maximum re-use, but of course the risk of using stuff that will disappear or be modified in uncontrolled ways, making your instance data invalid or undefined.

2.       Re-use existing vocabularies that are somehow deemed to be “good”, e.g. well-defined, well-maintained or owned by a trusted entity. For this, you need a set of criteria that determine what is “good” and what is not. This could be purely a set of local criteria, but you might also consider a set of globally accepted criteria. One of the ideas that I’ve heard was that there could be a Community of Vocabulary Owners that would agree on good practice in the form of a common set of maintenance and persistence policies, which could include the kind of commitments like the one between DCMI and FOAF (http://dublincore.org/documents/dcmi-foaf/). The advantage is that you would have some level of confidence that the vocabularies involved in this community would not disappear or break; the disadvantage is that such an approach takes time and effort in consensus building.

3.       Don’t directly re-use anything, but create parallel classes and properties in your own namespace with appropriate sameAs or subClass and subProperty declarations referring to other vocabularies as the alternative to re-use. I’ve heard the argument “Our project/service is going to be around longer than <fill in an organisation that maintains a vocabulary>” to argue for this approach. The advantage is that you’re not dependent on someone else’s policies or credibility, but I don’t think it will help the wider objectives of Linked Data. You’re moving the pain to the consumers who will need to resolve all these sameAs etc. relationships for incoming data to figure out that abc:title is really the same as xyz:title because they are both sameAs dc:title.

I had the impression that in the meeting it was suggested that W3C specifications may not want to refer to FOAF because it is outside of W3C which feels like going for option (3) – not re-doing FOAF of course, but in the sense of bringing existing FOAF under the W3C umbrella and associated policies. I hope that is not the general answer.

Makx.


Makx Dekkers
makx@makxdekkers.com<mailto:makx@makxdekkers.com>
+34 639 26 11 46