Re: How to use CDBase URIs in Content MathML?

Hi David,

  thanks for your explanations.  (I have taken into account that your reply
was rather personal than official, thanks for the clarification!  Same for
Robert's reply.)

Still, I'm not satisfied with the state of things and think that I have some
good reasons for thinking that it is suboptimal (see below), but OTOH I
appreciate your pragmatic workaround suggestions; I agree that that would
probably work in practice.

2010-07-12 16:14 David Carlisle <davidc@nag.co.uk>:
> cdgroup files dont currently allow aliasing[although see below] so 
> currently you would need to set up the alias externally, that is you 
> could have the mathml
> 
> 
> <math cdbase="http://example.com/mycdgroup.xml">
>    <apply>
>      <csymbol cd="foo1">bar</csymbol>
>      <csymbol cd="foo2">baz</csymbol>
>      <csymbol cd="foo3">fnord</csymbol>
>    </apply>
> </math>
> 
> with the mycdgroup.xml group file referring to renamed copies of the 
> original cd files (either actual copies or some kind of server redirect 
> set up at example.com.

OK, either that or the workaround below seems to work, but I think that
would increase the workload either on the side of the CD author (who has to
publish appropriate CDGroup files and redirects in addition to the CDs) or on
the side of the document/formula author (who has to set up CDGroup files in
case the CD authors haven't).

(Subjective rant: Keeping csymbol/@definitionURL or introducing an immediate
"strict" counterpart to it would have been so much easier.)

> the cdbase property is an optional openmath property in openmath2 and 
> not there at all in openmath 1, and while it's useful as a location aid 
> to find cd files I've never seen it used in practice to disambiguate cd 
> files of the same name.

OpenMath practice is lagging far behind the potential of recent web
technologies.  Please see my presentation at the OpenMath workshop for details
(http://cicm2010.cnam.fr/om/).  

Whether you want to do linked data applications (i.e. publish CDs at their
URIs) or not, I believe …

> The OpenMath CD review system has always assumed so far that any public CDs
> have unique names (hence the numbered names arith1, arith2 etc to allow
> easily naming variants)

… that OpenMath intends to be extensible in a decentral way, and that may
inevitably lead to CD name clashes.  Just suppose two researchers want to
write their own "arith"-like CD, look at openmath.org, notice that CDs up to,
say, arith6, already exist, name their CD arith7 each, and later on publish
the two arith7 CDs and notice the conflict.

More pointed:  If MathML doesn't require decentral, web-scalable extensibility
(which OpenMath gives, if used correctly), why didn't we stick with the
restricted vocabulary of Content MathML 2?

(Note, however, that redirects are not per se incompatible with linked data;
in the RDF world, some workarounds are also needed.)

> Arguably cdgroups do allow aliasing as the OM standard never actually 
> says that the CDName and CDURI in a cdgroup file have to match up,
> so if you were to take a forgving interpretation of
> 
> http://www.openmath.org/standard/om20-2004-06-30/omstd20html-4.xml#ssec_cdgroups

But doesn't the spec say that "the text occurring in the CDName element
corresponds to the name of a Content Dictionary in the CDGroup"?  I.e. by 

> <CDGroup>
> <CDGroupMember>
> <CDName>foo1</CDName>
> <CDURL>http://example.org/cd/foo</CDURL>
> </CDGroupMember>

we would be allowed to have a CD named "foo1" at the URL
http://example.org/cd/foo, but then, wouldn't that CD have to look like

<CD>
  <CDName>foo1</CDName>
</CD>

?

Cheers, and thanks,

Christoph

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