Re: attribute order (RE: Syntax Sugar and XML information models)

["Simon St.Laurent" <simonstl@simonstl.com>]

At 06:43 AM 3/30/01 -0800, Jeff Greif wrote:
>Even a non-validating parser must provide default values for the
>attributes not present, if such defaults are specified in the doc's DTD
>or schema.  If some element can have 1000 possible attributes, some mix
>of which is present in each instance, and you have a doc with several
>thousand of those elements, the parser will run very slowly if it cannot
>put the attributes in a hashtable when trying to fill in the defaults.

Is that hashtable necessary for each of the instance elements, or just for 
the default lookup table?  I'm not sure that you'd have to have a hashtable 
per element if your lookup table supported marking which attributes weren't 
necessary and then dumped a report of which remained.  (Reset per element 
instance.)

>Once it filled in even one default value, there is no longer an
>attribute order to be preserved for that element instance (do the
>defaulted attributes go at the beginning or end of the list, or
>somewhere else?)

A real and reasonable question, which raises yet another common issue - 
figuring out whether attributes were defaulted or not.  At least we have 
some answers on that one.  ('specified' in the Infoset, for instance.)

There's also the common question of whether attribute defaulting is a good 
idea - RELAX says no, quite plainly.  Even if you accept it as useful, I'd 
be curious to know whether defaults genuinely have no order, or whether 
that's something we just haven't really thought through.  DTDs certainly 
have internal order, for instance, and even something of a cascade on ATTLISTs.


Simon St.Laurent - Associate Editor, O'Reilly and Associates
XML Elements of Style / XML: A Primer, 2nd Ed.
XHTML: Migrating Toward XML
http://www.simonstl.com - XML essays and books