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

[Elliotte Harold <elharo@metalab.unc.edu>]

Dave Pawson wrote:
> Elliotte Harold wrote:
> 
>>
>> I would suggest a putative XML 2.0 follow roughly Tim Bray's 
>> skunkworks proposals. Specifically I'd like to see:
>>
> 
>> 3. Expand the list of predefined entity references to include what's 
>> defined in HTML and MathML.
> 
> Elliotte. You mutter about Unicode changes/stability, just how
> stable is this set of entities?

Reasonably. We'd just need to agree that we won't have every entity 
everybody needs; just a reasonable set that a lot of people need.

The current mess is caused by insisting as a moral imperative that we 
support every character anyone might ever want rather than accepting a 
plausible subset that fulfills somewhere between 99.9% and 100% of what 
people actually do need. Absolute positions that insist on fulfilling 
every conceivble use case have two problems:

1. They don't actually work. You can never cover everything.
2. Specs become too encumbered with corner cases to be plausibly 
implementable and understandable.

> Could this be handled in the same way that you propose for Unicode 
> version attributes?
> I.e. define them outside of xml 2.0 and reference the 'set' in use
> with the version of xml?

No, because then we'd need to define some spec for listing the entities. 
  The spec for Unicode properties already exists.

Perhaps we could simply allow undefined entities not to be a 
well-formedness error. (For a non-DTD-reading parser that's already the 
case) and allow different schema languages to define entities in their 
own ways; but I'd find that not especially interoperable and I'd prefer 
to avoid it.

We have 10+ years of experience now to know that very few people bother 
to define new character entities. Consequently I'd prefer to just leave 
that capability out of a smaller, simpler XML 2.0. There's always 
decimal and hexadecimal character references if you need it.

(Speaking of decimal character refs, that's probably something else we 
could drop in the name of simplicity. Hex is what almost everyone uses.)



-- 
Elliotte Rusty Harold  elharo@metalab.unc.edu
Java I/O 2nd Edition Just Published!
http://www.cafeaulait.org/books/javaio2/
http://www.amazon.com/exec/obidos/ISBN=0596527500/ref=nosim/cafeaulaitA/