Re: Let's get some principles nailed down

Simon St. Laurent writes:

>> Ouch.  That opens up whole new cans of worms.

I'm sorry if it does. as that was not my intention.

I presume we agree that applications should not attribute significance to 
the difference between:

        <e attr="1"/>

and

        <e attr='1'/>

Infoset is a good normative expression of a data model that shows why they 
are the same.   I don't think Tim's original reference to XML syntax 
captures the equivalence. 

>> definitions based on the Infoset rather than serial forms
>> sound like a wonderful excuse for organizations to define 
>> specs based on "open" XML and then turn around and use 
>> proprietary binary serializations.

That's not what I intended.  I would deal with your concern regarding 
proprietary binary with a separate recommendation along the lines of:

"XML 1.x SHOULD be used as the representation when XML is to be stored or 
communicated in byte- or character-serial form." 

Note, however, that I am intentionally supporting the use of mechanisms 
such as SAX and DOM as a means of communicating XML between applications. 
SAX and DOM are Infosets, but they are not XML 1.x (no angle brackets, no 
single or double quotes.)  I think that having one application or 
subsystem offer information in the form of a DOM or SAX stream, for 
consumption by another application (I.e. with no intermediate 
serialization to < >), is a fine usage pattern and should not be 
discouraged.  This pattern is commonly implemented by tooling such as XSL 
processors (Xalan has supported it for years), and is essential for 
performance, particularly of server-side XML applications. 

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------

Received on Friday, 22 November 2002 16:39:53 UTC