OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

 


 

   Re: [xml-dev] XML-with-datatypes (was....)

[ Lists Home | Date Index | Thread Index ]

On Fri, 2005-10-14 at 09:54 -0500, Peter Hunsberger wrote:
> On 10/14/05, Bullard, Claude L (Len) <len.bullard@intergraph.com> wrote:
> >
> 
> <snip/>
> 
> >
> > Do you think Vladimir's proposal to add to the xml: namespace to signify
> > stronger typing
> > is worth the risk of semantic expansion of that namespace?  As I said, I
> > don't see the
> > benefit of welding a strong typing mechanism to the core when it can be done
> > in the
> > application and there are variations on doing that (XSD builds in
> > primitives; RELAX
> > takes the bolt-in).  Optionality isn't a defense.  People trip on it just as
> > they
> > trip over the XML prolog (per Eric's blog on Lawyers Shouldn't Type XML).
> 
> A has been been observed, it seems that this thread is 80% perma but
> it amazes me that it's been over two years since this comment:
> 
> | Back in the days when I had time to hang out on the xslt list I found
> | myself giving a use case where strong typing would help us.  Now-a-days,
> | I've worked around it so much I no longer want it.  Essentially, we can
> | annotate a node from the back end with a type attribute and be done with
> | it once and for all; pretty much everything we ever needed to do with
> | types is now possible.
> 
> [1]
> 
> I'll make the observation, that this is still true, but I don't want
> just a single type attribute and I want to be able to define my own
> semantics for it.  My reasoning is as follows:
> 
> - If your data is travelling outside of a single well controlled
> domain then you either have to somehow standardize on a well defined
> type hierarchy or you have to allow for polymorphism on the types
> attached to any given element.
> 
> - Well defined type hierarchies may be possible but the effort to
> create them seems to be exponentially related to the number of users
> so their generality comes with a high cost (ie; XSD).
> 
> - Allowing each domain to attach a type that is semantically
> meaningful to them allows me to skip the cost of standardization and
> builds a loosely coupled type ontology for me at a much lower cost. 
> We can now discover that domain A has a "enrollment-date" that is
> somehow related to domain Bs "date-on-protocol" but we don't have to
> agree a-priori on which of these two terms will be used to define the
> type of a given element, (or exactly what they mean).
> 
> So, attributes it is, but ad-hoc attributes, and no W3C reserved
> namespaces unless we get some kind of uber-namespace (and I don't want
> to go there).
> 
> --
> Peter Hunsberger
> 
> [1]: http://lists.xml.org/archives/xml-dev/200306/msg00317.html

Well put.  I've written my own ideas and explorations on this theme in
several articles on IBM developerWorks:

http://copia.ogbuji.net/blog/2005-07-20/Thinking_X

And BTW, it's great to see Walter Perry back, weighing in on this topic
again.  Most of my sensibility of the importance of ad-hoc semantics in
XML derives ultimately from my reading his messages and articles.


-- 
Uche Ogbuji                               Fourthought, Inc.
http://uche.ogbuji.net                    http://fourthought.com
http://copia.ogbuji.net                   http://4Suite.org
Articles: http://uche.ogbuji.net/tech/publications/





 

News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS