W3C home > Mailing lists > Public > public-rdf-comments@w3.org > August 2013

Re: More clearly warn that "generalized RDF" is non-standard

From: David Booth <david@dbooth.org>
Date: Sun, 04 Aug 2013 17:07:59 -0400
Message-ID: <51FEC2AF.9050104@dbooth.org>
To: Sandro Hawke <sandro@w3.org>
CC: Pat Hayes <phayes@ihmc.us>, David Wood <david@3roundstones.com>, public-rdf-comments <public-rdf-comments@w3.org>
I would be okay in principle with including the definition somewhere 
with appropriate warning, but since the exact wording and context is so 
critical to its interpretation and impact, I would have to see the final 
verbiage to say for certain.

I do agree that the notion of "generalized RDF" is useful, but I don't 
think the W3C should encourage or promote its use.

I think the most appropriate place for the definition of "generalized 
RDF" would be in the RDF Semantics document, for two reasons: (a) that's 
where the notion is actually used (to simplify entailment rules); and 
(b) that would give it less prominence and hence reduce the likelihood 
that someone would think it is a form of standard RDF.

David

On 08/04/2013 07:57 AM, Sandro Hawke wrote:
>
>
> Pat Hayes <phayes@ihmc.us> wrote:
>> The terminology is already out there, eg
>> http://www.w3.org/TR/owl2-profiles/ is a Rec from 2012 which
>> defines and uses the idea (see
>> http://www.w3.org/TR/owl2-profiles/#Reasoning_in_OWL_2_RL_and_RDF_Graphs_using_Rules),
>>
>>
and  http://www.w3.org/TR/rif-rdf-owl/ is a 2013 Rec which has been a
>> Candidate Rec since 2009
>
> Actually a REC since 2009.   Otherwise I agree with all this.     :-)
> Hopefully we can add a little text to RDF Concepts so others
> understand this before starting a thread on the comments list.
> Assuming that text is okay, would you be satisfied, David Booth?
>
> - Sandro
>
> also defines and uses it.  (see
>> http://www.w3.org/TR/rif-rdf-owl/#Syntax_of_RIF-RDF_Combinations ).
>> The second one also has this in its "end note" (my emphasis
>> added):
>>
>> "Generalized RDF graphs: Standard RDF graphs, as defined in
>> [RDF-Concepts], do not allow the use of literals in subject and
>> predicate positions and blank nodes in predicate positions. The
>> RDF Core working group has listed two issues questioning the
>> restrictions that literals may not occur in subject and blank nodes
>> may not occur in predicate positions in triples. ***Anticipating
>> lifting of these restrictions in a possible future version of
>> RDF***, we use the more liberal notion of generalized RDF graph. We
>> note that the definitions of interpretations, models, and
>> entailment in the RDF Semantics document [RDF-Semantics] also apply
>> to such generalized RDF graphs."
>>
>> Apparently their anticipation was misguided, but I can attest that
>> almost any expressive extension of RDF would be easier and more
>> naturally defined on generalized RDF than standard RDF, regardless
>> of its lack of normativity. So it *will* get used. I don't think we
>> should discourage or encourage, just state the facts about what is
>> normative. Discouraging the use of generalized RDF amounts to
>> discouraging the use of RIF and OWL2 and JSON-LD and discouraging
>> the use of complete RDF reasoners, among other foolish things to
>> try to discourage. We are in the minority on this issue.
>>
>> I think DavidB's concerns might be alleviated if we could call it
>> by a different name, because "generalized RDF" does sound like a
>> kind of RDF. But this name is already in use, so to change it now
>> might cause more confusion. As usual, writing standards is a
>> lose/lose situation.
>>
>> Pat
>>
>>
>> On Aug 2, 2013, at 8:22 AM, David Wood wrote:
>>
>>> Yes, I concur with Sandro.  The RDF WG inserted the "generalized
>>> RDF"
>> description for a good reason.  It is there to allow for alignment
>> with JSON-LD and any future implementations or formats that cannot,
>> for good technical reasons, limit their possible parsings to
>> standard RDF.
>>>
>>> That does not mean, however, that we should encourage
>>> "generalized
>> RDF" at any point.  Its use should be strongly discouraged in
>> implementations and where that is impossible, as with JSON-LD, then
>> its /social/ use should be strongly discouraged, as with JSON-LD.
>> Hence, I think we should put in the stronger wording but leave the
>> concept in place.
>>>
>>> Chair and editor hats simultaneously "on" and "off".  Take that,
>> Erwin Schrödinger!
>>>
>>> Regards, Dave -- http://about.me/david_wood
>>>
>>>
>>>
>>> On Aug 2, 2013, at 09:15, Sandro Hawke <sandro@w3.org> wrote:
>>>
>>>> Speaking just for myself, I think this is going too far.   I
>>>> think
>> it does a service to the community to define the term "generalized
>> RDF" in RDF concepts, since (1) it's used in at least two of our
>> specs which we'd rather not have depend on each other (JSON-LD and
>> RDF Semantics), and (2) it's something people come up with on their
>> own anyway, and this way we tag the discussions about it.  It's
>> hard to evolve or extend a standard  interchange format, but the
>> best hope for doing so is to have everyone who wants to add some
>> feature add it in the same way and talk about it the same way.   By
>> defining "generalized RDF" in RDF Concepts, I think we're doing
>> that.
>>>>
>>>> All that said, I think it would be a good idea to add something
>>>> like
>> the warning note you propose, and perhaps some of the explanation
>> I just provided.   That is, roughly: Generalized RDF is not
>> standard RDF, but it can be useful and is reasonable to use among
>> systems which have all agreed to use it.  If you try to send it to
>> systems which have not agreed to use it, it won't work.
>>>>
>>>> -- Sandro
>>>>
>>>> On 08/01/2013 04:17 PM, David Booth wrote:
>>>>> I've been thinking further about this, and I have another
>>>>> more
>> radical suggestion.
>>>>>
>>>>> It seems to me that including even an informative definition
>>>>> of
>> "generalized RDF" in the RDF spec substantially increases the risk
>> that someone may mistakenly believe that "generalized RDF" is some
>> form of standard RDF, when it is not. It is an extension of RDF
>> that does not conform to the RDF standard. Hence it is all the more
>> important to visibly warn readers about the use of generalized
>> RDF.
>>>>>
>>>>> Actually, the more I think about it the more I am convinced
>>>>> that
>> the inclusion of the definition of "generalized RDF" in the RDF
>> spec **at all** is a big mistake, because it substantially
>> increases the risk that someone may mistakenly believe that
>> "generalized RDF" is some form of standard RDF, when it is not.
>>>>>
>>>>> Thus, my second suggestion is to entirely remove the
>>>>> definitions of
>> generalized RDF triple, graph and dataset from the RDF Concepts
>> document
>>>>>
>> https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#section-generalized-rdf
>>
>>>>>
>>>>>
>>
If the RDF Semantics document needs to define the notion of
>> generalized RDF to simplify the semantic rules, then I guess a
>> definition could be included in that document, *with* a big fat
>> warning saying that this definition is included only to simplify
>> the specification of the formal semantics, and does not constitute
>> a part of the RDF standard.
>>>>>
>>>>> David
>>>>>
>>>>>
>>>>> On 08/01/2013 11:20 AM, David Wood wrote:
>>>>>> Hi David,
>>>>>>
>>>>>> I acknowledge your comment and your concern.  I
>>>>>> *personally* agree
>> with
>>>>>> you that we need to carefully word this section of RDF
>>>>>> Concepts.
>>>>>>
>>>>>> The next RDF WG meeting that I will be able to attend is
>>>>>> 21
>> August,
>>>>>> so I will put this on the agenda for that meeting.
>>>>>>
>>>>>> Regards, Dave -- http://about.me/david_wood
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Aug 1, 2013, at 10:28, David Booth <david@dbooth.org>
>>>>>> wrote:
>>>>>>
>>>>>>> Section 7 defines the notion of "generalized RDF",
>>>>>>> triples and
>> datasets, but does not adequately warn that "generalized RDF" is
>> non-standard. Case in point: this has already led to some
>> discussion in the JSON-LD group about whether "generalized RDF" is
>> a form of standard RDF.
>>>>>>>
>>>>>>> I suggest rewording section 7 to the following, using a
>>>>>>> "NOTE"
>> call-out:
>>>>>>> [[ <p>It is sometimes convenient to loosen the
>>>>>>> requirements on <a>RDF triple</a>s.  For example, the
>>>>>>> completeness of the RDFS entailment rules is easier to
>>>>>>> show with a generalization of RDF triples.   </p>
>>>>>>>
>>>>>>> <p>A <dfn>generalized RDF triple</dfn> is an RDF triple
>>>>>>> generalized so that subjects, predicates, and objects are
>>>>>>> all allowed to be IRIs, blank nodes, or literals. A
>>>>>>> <dfn>generalized RDF graph</dfn> is an RDF graph of
>>>>>>> generalized RDF triples, i.e., a set of generalized RDF
>>>>>>> triples.  A <dfn>generalized RDF dataset</dfn> is an RDF
>>>>>>> dataset of generalized RDF graphs where graph labels can
>>>>>>> be IRIs, blank nodes, or literals.</p>
>>>>>>>
>>>>>>> <p class="note" id="note-generalized-rdf"> Any users of
>>>>>>> generalized RDF triples, graphs or datasets need to be
>>>>>>> aware that these notions are non-standard extensions of
>>>>>>> RDF and their use may cause interoperability problems.
>>>>>>> There is no requirement on the part of any RDF tool to
>>>>>>> accept, process, or produce anything beyond standard RDF
>>>>>>> triples, graphs, and datasets. </p> ]]
>>>>>>>
>>>>>>> Thanks, David
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>> ------------------------------------------------------------ IHMC
>> (850)434 8903 or (650)494 3973
>>
>> 40 South Alcaniz St.           (850)202 4416   office Pensacola
>> (850)202 4440   fax FL 32502                              (850)291
>> 0667   mobile phayesAT-SIGNihmc.us
>> http://www.ihmc.us/users/phayes
>
Received on Sunday, 4 August 2013 21:08:27 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:59:36 UTC