[
Lists Home |
Date Index |
Thread Index
]
4/26/2002 12:57:39 PM, Tim Bray <tbray@textuality.com> wrote:
>Except for, there's no de facto or de jure way to exchange an infoset as
>an infoset.
Not over the Web, but there are "de facto" ways to exchange infosets
between applications -- SAX events and DOM trees come to mind
(although implementation differences are a problem ... you have to be
very strict about accessing the standard interfaces rather than the
implementation classes).
> This is why XML is normatively defined at a syntactical level.
> Without reliable data interchange, you have nothing. With it,
> you have the potential to build anything.
"You have nothing" seems a bit strong. I'm pretty sure we are going
to see non-angle-bracket syntaxes for more and more "XML" data,
rationalized by conformance to the InfoSet, and justified by
more efficient use of bandwidth, compatibility with legacy systems,
tighter integration with programming languages, etc. XML as we
know it will be the "reference syntax" when no format negotiation
is possible, but people may do reliable data interchange based on
an ad hoc understanding (or future standardization) of how to
generate SAX events from some other exchange syntax.
I guess this bleeds over into the InfoSet thread earlier, (not to
mention the "binary XML" perma-thread), but I don't know how
I feel about this... a "good thing" because it extends the reach
of XML outside its traditional constraints, or a "bad thing" because it
violates the bedrock principle of reliable data exchange at the syntax level?
And even if we don't like the idea, how would we go about convincing
the people out there who are (as Don Box hinted last week) presumably
developing SOAP-specific routers, caches, firewalls, etc. not to go there,
considering that SOAP 1.2 is defined at the InfoSet rather than syntax level?
|
