Query from the trout pond

For reasons I don't understand (something about a "slap XXX around with a 
large trout" meme in IRC-land), the recent WSA F2F discussions referred to 
contentious, un-answerable issues as "trout ponds" rather than "ratholes."
When we made lots of progress on Thursday and got through the agenda with
a couple of hours to spare, the chairs opened up the floor to discussions
of issues that we had previously gavel-ed down because they were so prone
to going into ratholes/trout ponds.  

One issue that came to the surface very quickly was the increasingly 
unmanageable complexity of the XML-related specs themselves.  Assertions
were made along the lines of "only 7 people in the world really understand
XML Schemas" and "we couldn't get the gurus in the company to agree on what
seemed like a simple question of how to use namespaces for something."
I got the impression that this is causing significant productivity loss
and developer resistance in companies that have dug deeply into web
services.

Also, we agreed in principle that the WG would publish "position papers"
drafted by individual members or groups so long as they are approved by
a simple majority of WG members.  Since the subject of XML complexity is
one of my favorite hobbyhorses (c.f.  
http://www.xml.com/pub/a/2001/05/02/champion.html ) 
I may noodle on this a bit and draft a position paper on the implications
of all this for the web services architecture.

So, I'd like to reopen this trout pond a bit for discussion to get a 
better sense of :

a) do people agree that the complexity of the XML-related specs is a 
significant problem?

b) what particular aspects of XML (broadly defined) cause problems?  

c) would a "profile" of XML that deprecated DTDs and PIs but added
the features that SOAP/WSDL depend on such as namespaces and schema
types be useful? (Hint: processors could be optimized for this profile
and not have to support all the "cruft" that would presumably never
be used in a web services environment).

d) what very frequently used features of other specs might be put into the
XML Core to avoid the baggage of another spec?  For example, there
are proposals floating around to put the notion of "ID-ness" into the 
core (with an xml:id attribute, for example) so that one can reference
elements by their id without having to declare this in a DTD (which 
obviously wouldn't work in a SOAP context) or with a schema (which might
be overkill).

Received on Saturday, 16 November 2002 16:24:33 UTC