[Om3] RDF and some JSON [Re: [Om] Kicking off an OpenMath2+Process (Standard Enhancement)]

[Christoph LANGE]

Dear all,

[note: I'm now posting to om3 only]

first a comment on Manfred's proposal, then another one.

02/13/2011 07:10 PM Manfred Riem:
> I would propose to support more than the 2 encoding strategies.
> e
> One particular that comes to mind is a JSON encoding format.

I strongly support this idea.  My reason for supporting this is that
JSON is gaining more and more popularity as a lightweight interchange
format for web services.  The "X" for "XML" in AJAX is actually largely
obsolete, as XML is often considered too heavy for quickly sending
around small snippets of data between applications that have a common
understanding of OpenMath anyway.

Note that I not only support JSON for OpenMath objects, but also for
CDs.  (Well, via the "meta" CD, CDs can be represented as objects, but
the point of a JSON encoding is not that it just somehow encodes some
data, but that it does that in a way that is both intuitive and
space-efficient.)

My actual proposal, which I am happy to prepare as a formal SEP, is
something that might be called an "RDF encoding" in the context of this
discussion.  Here, the intended audience are not applications that know
OpenMath well, because for them XML, binary, or JSON are more suitable,
as they are less verbose.  But the audience is [web] applications that
do not know OpenMath but would still like to understand a bit about
mathematics, or applications whose developers do not want to hard-code
all OpenMath support into their implementations after having read the
specification.  Here, I am mainly thinking of an RDF vocabulary (a.k.a.
ontology) that represents the abstract CD model, and some of you know
that I have a complete proposal for that (chapter 3.2.3 of
https://svn.kwarc.info/repos/swim/doc/phd/phd.pdf), which we might use
as an input for whatever kind of standardization process.

On representing OpenMath _objects_ in RDF, my point of view is that
breaking down ordered n-ary trees into RDF triples is not a reasonable
thing to do.  However, I am aware of existing proposals to do it
nevertheless.  The most reasonable way IMHO is leaving objects in a more
suitable encoding (e.g. XML), while optionally redundantly modeling some
of their relevant properties (e.g. their type, or the root
operator/constructor, as has been done in MONET), in RDF.

In parallel to pushing that proposal, I will work on the promised
publication of the OpenMath standard CDs as RDF linked open data (see
http://arxiv.org/abs/1006.4057) and interlinking them with other
mathematical datasets.  I am currently resuming these activities and
preparing a CICM submission on that.  But that's just to give you some
background.  I consider that independent from any OpenMath 2+/3
standardization activities.

Finally note that there is a semi-standard way of encoding RDF graphs in
JSON, but by the same argument as above, going that route would miss the
point of a JSON encoding for _OpenMath_.

If you have any informal feedback so far, please let me know; otherwise
stay tuned for a more structured write-up.

Cheers, and thanks,

Christoph

-- 
Christoph Lange, Jacobs Univ. Bremen, http://kwarc.info/clange, Skype
duke4701
Semantic Publication workshop, May 29 or May 30, Hersonissos, Crete, Greece
Submission deadline February 28, http://SePublica.mywikipaper.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
Url : http://openmath.org/pipermail/om3/attachments/20110214/e3a8b7e1/attachment.pgp