Re: [xml-dev] The <any/> element: bane of security or savior of versioni

[noah_mendelsohn@us.ibm.com]

For what it's worth, I've slowly come to believe that the general case for 
forwards compatibility is not necessarily that new content is completely 
ignored by older applications, but that there be some default rule for 
dealing with it.  "Must ignore" is just one example of such a rule.  It's 
also perfectly reasonable, and often useful, to suggest that such 
additional content is to be stored along with other data, perhaps that it 
is to be printed using some default formatting rules, etc.  Note that in 
HTML as currently deployed, even elements not named explicitly in the 
spec, say <banana>, can be styled using CSS.   One point of view is that 
HTML+CSS therefore contains all such elements in its language.  Another is 
that <banana> is extension content, but with the rule that it still shows 
up in the DOM, still is scriptable, still can be styled via CSS, etc.  So, 
I've come around to the view that Must Ignore is just an extreme case of a 
default processing rule.  Note that if your very first step in receiving a 
V2 message in a V1 processor is to strip out everything that's new, you 
can't do things like digital signatures on it.   So my suggestion is that 
for forwards compatibility, early versions of a language need a very well 
chosen default processing (or default interpretation) rule:  whether the 
best rule is Must Ignore, and whether the rule can be implemented by a 
transform will depend on the circumstances and the requirements.

Noah

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------