W3C home > Mailing lists > Public > www-tag@w3.org > September 2004

Re: ACTION NW xmlChunk-44: Chunk of XML - Canonicalization and equality

From: Norman Walsh <Norman.Walsh@Sun.COM>
Date: Tue, 07 Sep 2004 10:49:46 -0400
To: www-tag@w3.org
Message-id: <87y8jmqgl1.fsf@nwalsh.com>
/ ht@inf.ed.ac.uk (Henry S. Thompson) was heard to say:
| Norman Walsh <Norman.Walsh@Sun.COM> writes:
|
|> I don't see why the comparison function should be doing validity
|> checking. You have an infoset. An infoset is a bag of properties. I
|> don't care where you got it from or how you constructed it.
|
| I absolutely agree -- you misunderstood one aspect of my email, see below.
|
|> | I hope that's not the answer, in which case I'd be interested not only
|> | in a specific explanation of why the [in-scope namespaces] EII
|> | property was not included, but also in the more general question of
|> | the implicit suggestion above that "Everyone knows what 'well-formed'
|> | means when applied to infosets" and "It doesn't make sense to define
|> | equivalence such that a well-formed infoset can be equivalent to a
|> | non-well-formed infoset."
|>
|> I hope my explanation above goes some way to expressing how I feel
|> about the answer to those questions.
|
| Not really -- you haven't explained why it's OK for two infosets to be
| equivalent despite one being 'well-formed' and the other not -- I like
| to think of members of equivalence classes as being inter-substitable
| for most important purposes, and serialisability is pretty
| important. . .

Unless the infoset you're serializing was constructed by an XML parser
From a serialized document and has been unchanged since its
construction, I think it would be poor design for a serializer to
assume that the [version] property on the document information item
was any more than a suggestion.

| This is _not_ to say that the equivalence test should check some
| definition of well-formedness, but rather that all (most) information
| item properties on which well-formedness depends should be included in
| the domain of the equivalence test.

And I think that the version property would overly constrain the test.

I think "<Xfoo/>" and "<Xfoo/>" are the same even if "X" is not a name
start character in a 1.0 document and the infoset for the first
occurance of "<Xfoo/>" in this sentence is labeled (incorrectly) as a
1.0 document.

                                        Be seeing you,
                                          norm

-- 
Norman.Walsh@Sun.COM / XML Standards Architect / Sun Microsystems, Inc.
NOTICE: This email message is for the sole use of the intended
recipient(s) and may contain confidential and privileged information.
Any unauthorized review, use, disclosure or distribution is prohibited.
If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.

Received on Tuesday, 7 September 2004 14:50:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:28 GMT