Re: [xml-dev] Ten Years Later - XML 1.0 Fifth Edition?

[noah_mendelsohn@us.ibm.com]

At least one reader of the attached noticed that I'd been somewhat sloppy 
in my use of the term "content".  I hope it's clear that I did not mean to 
imply that 5th edition allows for new "content" as distinct from markup, 
but rather in the more informal sense that the set of character sequences 
that are well-formed for 5th edition is, I believe, a superset or at least 
mostly a superset of that allowed by earlier editions.  I'm sorry for any 
confusion caused.

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








Noah Mendelsohn
02/20/2008 12:09 PM

        To:     Robin Berjon <robin@joost.com>
        cc:     elharo@metalab.unc.edu, XML Developers List 
<xml-dev@lists.xml.org>
        Subject:        Re: [xml-dev] Ten Years Later - XML 1.0 Fifth 
Edition?


Robin Berjon writes:

> The bottom line is this: it doesn't really matter. Or rather, of 
> course specifications have to be reliable over time, but they 
> certainly do not need to be perfectly reliable — just reliable enough 
> to be trustable by pragmatic companies.

Without commenting on the many other interesting issues on the table in 
this thread, that's just too simple.  I believe it really depends on the 
context and the application, and XML is used for many, many things.  One 
can easily imagine mission critical applications in which agreement on 
content is with reference to XML 1.0.  Note that SOAP 1.2 does just this 
to ensure that all SOAP nodes can deal with exactly the same content.  So, 

maybe the downstream code isn't robust against the new characters, because 

it's understood that they are filtered early, perhaps by an XML parser, 
perhaps by some other XML-aware code.  Maybe the parser gets updated to 
handle XML 1.0 5th edition and someone doesn't notice the dependency.

Can you argue that in a carefully run network these errors will be caught 
before causing trouble?  Yes, I think so.  Does reusing the XML 1.0 
"brand", which until now has been very stable as to what content is 
required significantly raise the risk that someone somewhere will be 
processing content that their code wasn't designed to deal with or tested 
against?  I think so. 

You are right that there are many situations in which "reliable enough" 
and "pragmatic" are the right bars to set.  I think it's too facile to say 

that there aren't others who reasonably rely on a specification like XML 
1.0 to be much more stable than that.  I'm not entirely sure, but I really 

do think both sides of the question need to be explored quite 
dispassionately.  My own personal leaning (though not necessarily my 
employer's), is that 5th edition is probably on balance a bad step and a 
bad precedent.  Interestingly, I could probably live with an XML 1.2 that 
allowed just the same content as 5th edition, and that made the XML 
declaration at the beginning optional, as it is in 5th edition.  The key 
difference is that anyone who says that they're depending on XML 1.0, or 
who asks for an XML 1.0 parser reliably gets the old rules.  If someone 
wants the new rules, they ask for an XML 1.2 parser, or a parser running 
in "1.2 mode".  Having a stable name for your specification is as 
important as determing what's on the wire.  I lot of my problem with 5th 
edition is that it use the XML 1.0 brand for both the old and the new 
content, which makes it very tricky to specify what you want when.

Noah


[1] http://www.w3.org/TR/soap12-part1/#soapenv

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