- From: Paul Vincent <pvincent@tibco.com>
- Date: Fri, 13 Mar 2009 05:02:27 -0700
- To: "Sandro Hawke" <sandro@w3.org>, <kifer@cs.sunysb.edu>
- Cc: "Chris Welty" <cawelty@gmail.com>, "Axel Polleres" <axel.polleres@deri.org>, <public-rif-wg@w3.org>
May I suggest that PS concentrates on the simpler (?) needs of BLD / Core rather than PRD? FWIW: it seems likely that in any <BRE PR language> --> <RIF PRD> mapping, the RIF PRD version is going to effectively obfuscate the original meaning of the production rules; thence a presentation syntax of the RIF PRD version is only going to be used for / have a use case of "explaining the mapping" (i.e. not for general reading and writing of PR rules). [IMHO: we will likely need a RIF PRD-OO (i.e. object oriented) that is a sub-dialect of RIF PRD before we get widespread support from the BRE vendors; a Presentation Syntax for RIF PRD-OO would probably require at least a set of extensions for the RIF PS to be truly presentable]. My 2c... Paul Vincent +1 650 206 2493 / mobile +44 781 493 7229 > -----Original Message----- > From: public-rif-wg-request@w3.org [mailto:public-rif-wg- > request@w3.org] On Behalf Of Sandro Hawke > Sent: 13 March 2009 05:12 > To: kifer@cs.sunysb.edu > Cc: Chris Welty; Axel Polleres; public-rif-wg@w3.org > Subject: Re: Presentation Syntax > > > Wow, do I not want us arguing about presentation syntaxes. :-( > > What I'm picturing now is having one "RIF Presentaiton Syntax" (RPS), > which is a superset of the current BLD Presentation Syntax, where > everything added is either needed for PRD or is trivial syntactic sugar > (like infix operators and simple datatype literals). > > The process I imagine is this. I'll add anything that: > (1) I can implement relatively easily > (2) someone wants > (3) no one objects to. > > If anyone objects, that's it. No more discussion in meetings. You can > take it off-line with them if you really want, and if you change their > mind, fine. But you're encouraged to just let it drop. > > I also picture folks forking RPS. I hear Michael objecting to Turtle > syntax, so fine, it stays out. But that doesn't stop someone from > creating RPS-T or whatever. And if there's an implemented translator > to > XML, they can use if for test cases (which is something we already > decided). On the other hand (and this is important), it doesn't need a > formal spec, and it wont actually be mentioned in any of our documents > -- just used in the test case repository, and maybe the on-line demo. > > Good enough? Can we do this, and stop arguing about the PS? > > (When I have a next version implemented, I'll document exactly how it > differs from what's in BLD, for people to potentially object to, and > potentially add to. Until then, let's stop talking about the PS.) > > -- Sandro > > > On Thu, 12 Mar 2009 23:21:56 -0400 > > Chris Welty <cawelty@gmail.com> wrote: > > > > > > > > Michael, > > > > > > The presentation syntax is not for "exchanging concrete languages". > The XM > > L > > > syntax, which is not in dispute, is for that purpose. The > presentation syn > > tax > > > is pretty much only for *humans* who want to write (and read) rules > in BLD. > > > > It is for the **semantics** and therefore is normative. > > We also use it for examples etc., for which it is more than adequate. > > > > > Since you think BLD is useless, you shouldn't care what syntax > humans want > > to > > > use to write&read it. > > > > BLD is useless because it is inexpressive (where it matters). But it > is part > > of > > the future dialects that I care about. > > > > > You can use your personal syntax for the dialect that you > > > find useful, as long as it is 1-1 translatable to the XML syntax > there is n > > o > > > problem. > > > > > > In fact, the ability to have multiple presentation syntaxes seems > to me to > > > precisely *demonstrate* the advantages of RIF. In some sense, each > of thes > > e > > > syntaxes can be thought of as different languages which have very > good > > > interchange properties through RIF-XML. > > > > We are not in the business of defining multiple new languages. RIF > was suppos > > ed > > to be an exchange format among existing languages. If somebody wants > to defin > > e > > their own language for whatever use, they can do that and exchange it > through > > RIF. But there is no place for such a language within RIF itself. > > > > > Neither RDF nor OWL have suffered from multiple equivalent syntaxes > (though > > the > > > same arguments were made against them at the time), I see no > evidence that > > it > > > would bloat or harm adoption of RIF at all. Quite the opposite. > > > > I don't care if you write 20 new papers on 10 different alternative > syntaxes > > in > > the next 5 years. I care about the normative syntax that is used to > define th > > e > > semantics, especially the semantics of the framework. > > > > michael > > > > > > > -Chris > > > > > > Michael Kifer wrote: > > > > Axel, > > > > pls see my earlier response to Sandro. > > > > I believe that all these proposals are leading the group astray > and causi > > ng us > > > > to loose sight of what RIF is about. To me, RIF is not BLD and > not PRD, b > > ut a > > > > framework for dialects. BLD, IMO, is mostly useless. I see the > future of > > RIF in > > > > dialects that extend BLD. You are proposing to introduce > completely usele > > ss > > > > (IMO) bloat into what is supposed to be a framework for > exchanging concre > > te > > > > languages. > > > > > > > > michael > > > > > > > > > > > > On Thu, 12 Mar 2009 18:35:09 +0000 > > > > Axel Polleres <axel.polleres@deri.org> wrote: > > > > > > > >> Michael Kifer wrote: > > > >>> On Thu, 12 Mar 2009 16:44:10 +0000 > > > >>> Axel Polleres <axel.polleres@deri.org> wrote: > > > >>> > > > >>>> Michael Kifer wrote: > > > >>>>> I think the origin of this thread was your proposal to use > triples {a > > b c} > > > >>>>> instead of frames. This is unacceptable because frames are > not triple > > s. > > > >>>> SWC defines them as corresponding [2]: > > > >>>> > > > >>>> "This means that whenever a triple s p o is satisfied, the > correspondi > > ng > > > >>>> RIF frame formula s'[p' -> o'] is satisfied, and vice > versa[...]where > > > >>>> s', p', and o' are RIF symbols corresponding to the RDF > symbols s, p, > > > >>>> and o, respectively." > > > >>> I am saying: frames are not triples, but triples are frames. > > > >> That's ok. > > > >> > > > >>>>> You can have a[b->c d->e] etc. > > > >>>> That is not a problem, in PS I would likewise suggest to allow > not onl > > y > > > >>>> frame style but also turtle style abbreviations, cf. [1] i.e. > > > >>>> > > > >>>> { a b c; d e } etc. > > > >>> Hmm. You don't expect me to like that, do you? > > > >> There are many people who do like and regularly use that. > > > >> > > > >>>>> Second, {...} is not a good syntax because it means sets. > > > >>>> not in RIF. From that point of view, one could equally argue > "[...] is > > > > > >>>> not a good syntax because it means array or lists". So, again, > I don't > > > > > >>>> see the problem. > > > >>> Set notation is fundamental, and there has been demand for that > in use > > case > > > >>> examples. I didn't introduce that before because originally the > > > >>> presentation syntax was supposed to be abstract syntax. But now > when we > > are > > > >>> moving towards a concrete syntax this is on the agenda. > Aggregates are > > > >>> definitely on the agenda even if the set syntax won't make it. > > > >>> > > > >>>>> This is what I was > > > >>>>> planning to use for aggregates (I proposed this several > months ago in > > an email) > > > >>>>> and for shorthands like a[b->c b->d] == a[b-> {c d}] > > > >>>> I do not see this approved anywhere, as of yet. > > > >>> See above. > > > >>> > > > >>>> Alternatively, I suggest > > > >>>> for that case to use the Turtle style abbreviation [1] using > commata f > > or > > > >>>> separating multiple values, i.e. > > > >>>> > > > >>>> { a b c, d } > > > >>>> > > > >>>> or, in frame style: > > > >>>> > > > >>>> a[b-> c, d] > > > >>>> > > > >>>> My preferences is compatibility among W3C promoted languages > (RIF, RDF > > , > > > >>>> SPARQL). > > > >>> This is doomed to failure and is a waste of time. > > > >> Well, we are already there: it needs no time at all, if we just > accept > > > >> Turtle style patterns in { ... } as a notation for conjunctions > of > > > >> frames. So, no additional time needed, and no time wasted. > > > >> > > > >>>> FWIW, I don not necessarily propose to *replace* frames, but > to > > > >>>> allow Triple/Turtle style syntax alternatively at lease for > the sake > > > > > >>>> of this aimed compatibility. Would that be acceptable for you? > > > >>> No. Why duplicate and complicate? > > > >> It does not complicate anything in my opinion, and it is a > reasonable > > > >> alternative to frames style syntax with - for a W3C WG in the > Semantic > > > >> Web activity - a significant community behind. > > > >> > > > >>>> The SW community is becoming very used to this kind of > syntax and a > > > >>>> compatible PS would certainly be appreciated from that side. > > > >>>> > > > >>>> BTW, I would actually not expect too many people wanting to > mix Turtle > > > > > >>>> style and frame style syntax in the same PS document... > > > >>> Don't forget that RIF is an exchange syntax despite all the > talk about > > the > > > >>> presentation syntax. If you want to use turtle syntax then use > Turtle s > > yntax > > > >>> and translate it to (XML) RIF. > > > >> Just to clarify: Turtle is not a rules language, but it can > model the > > > >> part covering conjunctions of frames/triples. > > > >> > > > >> best, > > > >> Axel > > > >> > > > >>> cheers > > > >>> michael > > > >>> > > > >>> > > > >>>> best, > > > >>>> Axel > > > >>>> > > > >>>> 1. http://www.w3.org/TeamSubmission/turtle/#sec-tutorial, > Section 2.3. > > > >>>> 2. http://www.w3.org/2005/rules/wiki/SWC#RDF_Compatibility, > Section 3 > > > >>>>> michael > > > >>>>> > > > >>>>> On Thu, 12 Mar 2009 09:23:12 +0000 > > > >>>>> Axel Polleres <axel.polleres@deri.org> wrote: > > > >>>>> > > > >>>>>> Sandro Hawke wrote: > > > >>>>>>> Talking after the call today, Axel made a good case for > adopting th > > e > > > >>>>>>> other solution to the infix-operators problem, namely to > pay attent > > ion > > > >>>>>>> to whether there is a space between the "f" and the "(" in > "f(x)". > > So > > > >>>>>>> "f (x)" is two terms, and "f(x)" is one term. This is my > own > > > >>>>>>> preference, and what I implemented before the task force > telecon (w > > here > > > >>>>>>> I got talked out of it). One of Axel's considerations is > alignment > > with > > > >>>>>>> SPARQL and other RDF-related languages, in user training > materials. > > If > > > >>>>>>> we go with this kind of space-sensitivity, then we can have > both th > > e > > > >>>>>>> F-Logic and N3 styles available for expressing triples, so > you can > > write > > > >>>>>>> your exampes in either one. And you can use commas if you > want, bu > > t you > > > >>>>>>> can also leave them out. This means my parser should be > able to ha > > ndle > > > >>>>>>> all the existing examples and test cases... a pretty big > win, reall > > y. > > > >>>>>>> The one thing people can't do is write "f (x)" when they > mean "f(x) > > ". > > > >>>>>>> > > > >>>>>>> -- Sandro > > > >>>>>> In that context, note that in the proposed PS we hav space- > sensitivi > > ty > > > >>>>>> already anyways, for the '-' operator vs. dashes within > IRIs. So, I > > > >>>>>> think that additional space-sensitivity would be totally > acceptable. > > > >>>>>> > > > >>>>>> Axel > > > >>>>>> > > > >>>>>> > > > >> > > > > > > > > > > >
Received on Friday, 13 March 2009 12:03:18 UTC