xml-dev - Re: To Relax or not to Relax

[Rick JELLIFFE <ricko@geotempo.com>]

Terren Suydam wrote:
 
> Do you think it makes sense to use Relax now, as a transition between DTD-based
> validation and XML Schemas?

 
> Do you think Relax might ultimately be superior to XML Schemas?

Is the motorbike superiour to the motorcar?  It is like that Woody Allen
movie where his parents argued about whether the Atlantic Ocean was
"greater" than the Pacific. 

Now that XML Schemas is pretty much complete we can look at it and
figure out what it is good for.  

For datatyping, XML Schemas and RELAX are pretty much the same. I'd say
they are good if you have transnational data requirements or systems,
especially if you need to connect these to databases or to send data
between backend systems. 

But if you want to send your data in formats, you have to fall back to
calling all your datatypes strings, and using lexical typing. For
example, if you want to send dates as "03/32/00" you cannot say "this is
a date" but you can just say "this is a string.  (I regard this as a
flaw that could easily be recitifed by treating notations properly.)  

So it may be that XML Schemas fits into a world where data is sent
efficiently from a database in transnational formats and some
application layer makes it palatable for humans to use.  This approach
obviously suits large vendors of database and business systems (e.g.,
Oracle, IBM, SAP, MS, Sun) but does it really suit the needs of people
making non-business-related Schemas in XML?  Do the datatypes help XSL
or WAI or the i18n people?  Without pictures, are the datatypes suitable
for use in forms?

(And if XML Schemas is designed to be used with an intermediary
application between it and the user, why is there the need to allow one
generic identifier to identify elements of different types depending on
context: why is it important for the user to be able to reuse element
names but not important to use familiar data values?)

As far as the structures spec goes, both XML Schemas and RELAX are based
on grammars (where the parent or grandparent determines the child). 
They have different models of the importance of extendability (XML
Schemas) versus modularization (RELAX). These certainly impact the
managability of each, but I don't think that Murata-san or the W3C XML
Schema WG could warrant that one is universally superior to the other. 
You may find the RELAX core to be adequate (i.e., the subset of XML
Schemas that RELAX shares); we simply don't have enough experience to
know whether people for whom that common set is not enough are
adequately catered for by the full RELAX or the full XML Schemas.

I think the important difference between XML Schemas and RELAX has not
been commented on widely, but it is no co-incendence that Murata-san's
RELAX and my Schematron both do not make "information set
contributions"--indeed I found Murata-san's thoughts on this issue very
compelling and I would be very interested to know what he (or the people
who have taken up RELAX) currently thinks about this.

The issue boils down, perhaps, to whether it is desirable to invent a
new class of XML documents: ones which need to be "schema processed" in
order to be used. From a power and capability view, it would sure be
nice--we can have XPaths which include type awareness (and then
Schematron would inherit that!), XBase would know what strings were URIs
in documents, and so on.  But is this really creating an overly-layered
and bloated framework that will have performance problems over the WWW
and which will create many excuses for dialects?  

But this new class of XML documents is where we may be heading.  

I suspect that many XML-DEV readers might tend to favour lean-and-mean
technology for the web, and it has not been demonstrated yet that XML
Schemas is lean-and-mean.  That is why it is good that XML Schemas has
not gone directly to Recommendation, and it is why W3C should be
encouraged IMHO to have a Candidate Recommendation period of at least
several months; and most importantly it is why we should not regard it
as a defeat or victory if XML Schemas has to go back to the XML Schema
WG for revision at the end of the CR period.  

I hope the people in XML-DEV will try to be as constructive as possible
during this CR period: try XML Schema betas, see if they do what you
want...try RELAX..try Schematron even.  I think Murata-san has been
admirably constructive in this regard: in RELAX he has helped point out
which parts of XML Schemas may be regarded as core and which parts may
be regarded as peripheral.  And, most importantly, he has raised the big
issue of what processing architecture the XML WWW should have.  

Even though I, obviously, tend to think the whole grammar-based approach
(as used in all the schema languages) is a little bogus (I have a draft
of a new Schematron paper on this at
http://www.ascc.net/xml/resource/schematron/Schematron2000.html,
comments welcome), it is obvious that we are stuck with grammar-based
schema languages and that we will have to make the best of it. 


Rick Jelliffe

***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************