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

[Jonathan Robie <jonathan.robie@redhat.com>]

Elliotte Harold wrote:
> For the record, I would support an XML 2.0 that addressed these 
> issues, provided it offered enough benefits to outweigh its costs. The 
> problem with XML 1.1 and XML 1.0.5 are that they benefit *no one* and 
> impose huge costs on everyone. (I still have yet to meet one single 
> actual user who needs XML 1.1.)

I really dislike XML 1.1, and it has complicated our XQuery / XPath / 
XML Schema specifications to have to support it alongside XML 1.0.

To me, XML 5th Edition is a different beast, largely because it does 
*not* impose a huge cost on an implementer who uses Unicode libraries or 
language environments that support Unicode.  There's no need for an XML 
parser to try to interpret or understand the new characters, which are 
just passed through. If an XML parser correctly implements its 
non-support for these characters, it does so using code that tests these 
ranges specifically to reject such characters, so the impact would be to 
remove such  code. I've talked with one implementor only, but he said 
this would simplify his code and require very little time to implement.

And I don't think it imposes huge costs on users either, it just lets 
them use the characters in their character set.

> I would suggest a putative XML 2.0 follow roughly Tim Bray's 
> skunkworks proposals. Specifically I'd like to see:
>
> 1. Combine the namespaces, XML base, and xml:id specs with XML 1.0. 
> (Possibly XInclude, though I'm not sure about that one.)
> 2. Remove the internal DTD subset. Allow the DOCTYPE declaration to 
> point at schemas of various types and move all discussion of validity 
> to separate documents for different schema languages.
> 3. Expand the list of predefined entity references to include what's 
> defined in HTML and MathML.
> 4. Expand the name productions to include characters from Unicode 5, 
> but still forbid undefined characters, musical symbols and the like. 
> That is, follow the patterns of XML 1.0 rather than 1.1.
> 5. Ban the C0 and C1 control characters, except \r, \n, and \t.
> 6. Eliminate CDATA sections
> 7. Eliminate one of the quotes, either single or double, around 
> attribute values.
> 8. Remove attribute value normalization and all attribute types (at 
> least in the base spec)
>
> I'd be willing to compromise on a lot of this, by the way. I can live 
> ith CDATA sections and single quoted attribute values, even if they 
> make life tougher for parser writers.
>
> That would be a simplified XML worth supporting. However what XML 1.1 
> gave us and XML 1.0.5 is now proposing is of no value to anyone, and 
> imposes massive costs on everyone. It is a simply bad idea and a bad 
> proposal, even irrespective of the abuse of the errata process.

I think this would be *much* more disruptive than XML 1.0 Fifth Edition, 
since some XML parsers would suddenly be unable to read much of the 
existing corpus of XML. And the process of coming to agreement on which 
features are in or out, and which other features people might like, 
would be a long and contentious one.

In the meantime, XML is 3 versions behind the Unicode standard.

Jonathan