At 09:05 PM 7/31/00 +0800, Rick JELLIFFE wrote:
> The infoset lets people know what information will be in the parsed XML,
> regardless of which lexical form was used. 

If I actually believed that was all the impact the Infoset would have, I'd
be far less worried.

I've been somewhat amazed to see how people all tend to have different
views of the amount of information that needs to be presented 'as the XML
document' to an application.  DTDs seemed to be the first to go, and I
think a lot of us miss them, even in application processing.  For a long
while, namespace prefixes were discarded - and then the Schemas and XLink
folks found them useful in attribute values.

John Cowan's character reference examples feel convincing to me - in
exactly the opposite manner of what he intended.

I'm not sure there is any  value to creating an abstraction that only
represents one 'processed' phase of what a syntax is capable of, when it's
the only 'blessed' abstraction.  I'd have found the Infoset far more useful
if it had started by representing the items actually available in XML
syntax and then defined the processes by which they get reduced to a
smaller set of items.

(Yes, I'm aware the Infoset has 'value' to those who find it meets their
needs - I'm also aware that it imposes a cost on those whose needs are not
me, simply by being the 'Recommendation' and blocking other approaches.)

Sadly, the Infoset seems stuck on one idea, and derailing moving
philosophical trains at the W3C appears to be damn near impossible.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
http://www.simonstl.com - XML essays and books