[Om3] CDSignatures/@cd, CDSignatures/@type: how to resolve?

[Christoph LANGE]

Dear James, dear David,

  let me reply to all of your mails at once, in reverse chronological order:

On Friday 09 May 2008 12:27:24 David Carlisle wrote:
> James
>
>   Sorry - poor wording on my part. What I meant to say is "what is meant by
>   'the standard arith1'?". As I see it, in the absence of an EXPLICIT
>   CDbase, the application is meant to know, either inherently OR by going
> to www.openmath.org, the semantics of the OMS.
>
> yes, agreed.

That sounds perfectly reasonable to me. But then we should mention in the
specification that, if no cdbase is declared in a document, and an application
does not have its own notion of a CDBase (as e.g. a mathematical database
might have), http://www.openmath.org/cd must be supported as a fallback, what
David called "default-default".

On Friday 09 May 2008 12:07:42 James Davenport wrote:
> On Fri, 9 May 2008, David Carlisle wrote:
> > JHD wrote
> >
> > > Something does seem to be wrong, but it may be my own ignorance.
> > > Let me ask a stupid question. When I say
> > > <OMS name="plus" cd ="arith1"/>
> > > I mean 'the standard arith1'. Can you refresh my memory on how that
> > > intention is implemented?
> >
> > I think historically it was implemented by whatever your system wanted
> > to implement. There are good reasons for that, OM systems are not all
> > web facing, or even web aware.

Sure, I didn't mean that every system must be able to access the web. But I
meant that if some system S references arith1#plus without giving a CDBase we
do not know what "arith1" is, whether it is the same CD as
http://www.openmath.org/cd/arith1.ocd, or whether it is just something that
happens to have the same name within S but contains completely different
definitions. That's also what RDF uses URLs for: not always for actually
making a connection to some web server and retrieving data, but mostly just
for giving things (here: symbols) unambiguous names.

> > however OM2 did add a sort of hint that
> > you might want to generate URI out of a CD and symbol name, and OM3 (and
> > certainly MathML3) make that association rather more strongly.

So any OpenMath-aware application, be it web-aware or not, should still be
aware of the fact that the full, unambiguous name of the "plus" operator as it
is officially defined in the openmath.org CDs is
(cdbase="http://www.openmath.org/cd", cd="arith1", name="plus"), or
http://www.openmath.org/cd/arith1.ocd, if the application chooses to use URLs
internally -- that's what I meant to say. Just that we make sure that
different systems that have their own local representation of CDs are actually
talking about the same thing when they say "arith1#plus".

> I don't understand (this may be because I don't speak RDF) …

RDF, explained in one sentence, is actually quite simple: you make
subject-predicate-object statements about things, and you reference these
things unambiguously by URIs.

> > > > Then @type="sts" would resolve to http://url/of/sts.ocd, and
> > > > @cd="foo"
> > >
> > > I would seriously hope not - I would like it to refer to 'the standard
> > > sts' as with "arith1" above.
> >
> > I don't think the CDbase should resolve automatically to the base URI of
> > the signature file.
>
> I'm not sure if we agree here, or not.

So might it help to have a CDSignatures/@cdbase attribute in order to
disambiguate? As said above, we could say that it is optional and defaults to
http://www.openmath.org/cd. But one more question is: Does it always make
sense to assume the same CDBase both for resolving CDSignatures/@type and
CDSignatures/@cd? For my own needs this questions is a hypothetical one, but
still, could there be a case where the type system CD is in a different place?

On Friday 09 May 2008 09:50:58 David Carlisle wrote:
> I suspect that the simplest (and perhaps even best) thing to do is just
> to specify that CD names are taken relative to the system's CDBase, and
> then extend the signature file schema (currently omcdsig2.rnc) to allow
> a CDBase element (as is already allowed in CDs) then to say that if you
> are reading a signature file, then  CD base is as specified in the file,
> or if it is not specified, some system-dependent knowledge, or a
> default-default of http://www.openmath.org/cd.

I completely agree, see also above.

On Friday 09 May 2008 07:17:17 James Davenport wrote:
> > signature dictionary to the CD it contains signatures for and to its type
> > system is established. The specification says that both the @type and the
>
> I assume the 'specification' means OM3

Both OM2 and OM3, in this case it's the same.

> > @cd attribute of CDSignatures are just _names_ of content dictionaries,
> > but how
> > should these be resolved? How would we find out their URL? If we're just
> > looking at a signature dictionary, we don't have any information such as
>
> This is a good question: 'does/should a signature file make sense "on its
> own" and, if so, how?'. You are saying 'Yes, but indeed how?'

As I'm OMDoc-biased I could of course imagine having the type signatures right
in the CDs instead of separate files. But for OpenMath I just accepted the
status quo that we have type signatures (and notation definitions) in separate
files. But, yes, if we have separate files, they must be able to unambiguously
refer to each other.

> Something does seem to be wrong, but it may be my own ignorance.
> Let me ask a stupid question. When I say
> <OMS name="plus" cd ="arith1"/>
> I mean 'the standard arith1'. Can you refresh my memory on how that
> intention is implemented?

As I said above, the spec does not yet mention that "the standard" means
http://www.openmath.org/cd.

> > Here is an example for resolving relatively to the signature dictionary.
> > I assume that's the common case, but then this default behaviour should
> > be mentioned in the specification:
> >
> > Let http://url/of/foo.sts be a signature dictionary as follows:
> >
> > <CDSignatures type="sts" cd="foo">
> >   ...
> > </CDSignatures>
> >
> > Then @type="sts" would resolve to http://url/of/sts.ocd, and @cd="foo"
>
> I would seriously hope not - I would like it to refer to 'the standard
> sts' as with "arith1" above. I do not know enough RDF to know how to
> achieve this effect, though.

This is not a question of RDF, just a question of defining the "base URI" (in
fact the CDBase) relatively to which @type and @cd are to be resolved.

In general, my SWiM point of view is: I do not have particular requirements
for getting my work done with OpenMath. I merely want to develop SWiM into an
editor that supports the way OpenMath users work. But as I'm using semantic
web technologies (and that is e.g. RDF) in order to do so, I must care about
making OpenMath compatible with the semantic web in places where it is not
yet.

Best,

Christoph

-- 
Christoph Lange, Jacobs Univ. Bremen, http://kwarc.info/clange, Skype duke4701



-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
Url : http://openmath.org/pipermail/om3/attachments/20080509/7e353777/attachment.pgp