RE: Options for dealing with IDs

No.  Len wants read on demand. One can't easily get around 
the network costs of two fetches to process one resource, 
so one has to insist that it be on demand but that is not
the same as optional.  

Otherwise and only once because this is the TAG list and 
this debate belongs on XML-Dev:

I am concerned about creating new functionality that replicates 
old functionality and a direction that leads to wholesale 
replacement of the old functionality without regard to 
those who depend on it.   The XML community in some part 
has an "XML Uber Alles; SGML die" mentality.  That will only 
be acceptable when XML does all the jobs SGML does at least 
as well as it does them.

Ultimately, we are on the path to ditching XML DTDs.  It 
may be by the typical W3C "one tiny step down the 
path" route, but that is where this is going.  So 
at each step, we will keep reexamining the impact of 
those steps on everyone on that journey.   DTDs are ugly 
for the programmer, but easy for the author.   XML by 
loading up the root with evermore PIsInDisguise is getting 
blisteringly hard to read and understand.  Chris does a 
good job presenting the options.  We have to do a good 
job figuring out who pays which bills per option.  xml: 
is the 'bag o' atts' approach.  It may be the best solution 
for some aspects of pure decoration properties, so I tend 
to support Chris.  But also, I tend to support Paul that 
such still belong in single specs until we get a better 
understanding overall of the wholesale direction of bag 'o atts 
on the complex XML framework and the more complex set of 
diverse users.  Tell me this will make life easier for the 
author AS WELL as the parser writer and XML programmer, 
and I'm a lot happier guy.

len

-----Original Message-----
From: Elliotte Rusty Harold [mailto:elharo@metalab.unc.edu]

Ultimately, the document is just a document. Len seems to think that 
we should make all processes read the DTD and accept what's written 
there. I'm going the opposite way. Each process applies the semantics 
to the document that make sense for it. This includes determining 
what is and isn't an ID.

Received on Wednesday, 8 January 2003 11:28:29 UTC