[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: UTF-8+names

• To: Mike Champion <mc@x...>
• Subject: Re: UTF-8+names
• From: Bill de hÓra <bill.dehora@p...>
• Date: Sun, 19 Oct 2003 15:36:25 +0100
• Cc: xml-dev@l...
• In-reply-to: <20031019034635.91162.qmail@w...>
• References: <20031019034635.91162.qmail@w...>
• User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4) Gecko/20030624

Mike Champion wrote:


> Tim's approach is taking a real, widespread problem
> and offering a clean, layered solution -- essentially
> a character encoding preprocessor -- rather than
> changing XML itself.

Exactly. It's a syntax macro (just like namespaces). So the 
*problem* of resolving entities hasn't gone away, specifically that 
  problem is one of getting your entities from one place to another 
and back again. Unless of course we all upgrade to use +names.

I'd love to know what SOAP-oriented folks think about it.


> Actually, a similar idea came up at the Binary Infoset
> workshop, to leverage/exploit the fact that the XML
> spec allows an open-ended set of encodings. This
> allows experimentation WITHOUT "corrupting" the core
> spec with support for local languages, stuff of
> interest mainly to mainframes, or more efficiently
> transmittable and/or lexable serializations.  

It also potentially hurts interop - there are costs and benefits to 
be weighed up.


> IMHO, it extends the Unicode encoding layer upwards to
> remove a wart in XML, not vice-versa.

I disgree :) It sems to be solving a wart (or a gap in the market) 
of Unicode transformations, not XML. Otherwise why would it be 
useful outside XML?

> 
> Anyway, I think this is a great idea, and I
> congratulate Tim for working it out and moving it
> forward.  

I like it at first glance, but the current draft is too vague. I 
suspect the impact of this encoding is more than Tim is giving 
credit for - so I'm not buying arguments from idiotic simplicity 
just yet.

I'd like to:

  o see the encoding name changed to "UTF-8+entities", the current 
name is rather vague.

  o see examples of escaped whitespace.

  o know whether <w&oacute;oops/> is a legal element name in this 
proposal.

  o hear a rationale, other than use outside XML, for choosing a new 
encoding to solve this problem, ie why not xml:entities="yes" or 
some other approach?

  o know whether the current MathML/HTML4 sets are sufficient; ie 
are we going to need to reversion this in couple of years to cater 
for ogham?

  o like elharo and Alessandro, I'm unconvinced about treatment of 
&, specifically, that it isn't being overloaded in some 
clever/sneaky way. Indeed, I'll claim that & /is/ being overloaded 
in some clever/sneaky until the next draft shows me otherwise.


Bill de hÓra
-- 
Technical Architect
Propylon
http://www.propylon.com

• Follow-Ups:
  • Re: UTF-8+names
    • From: John Cowan <cowan@m...>
  • Re: UTF-8+names
    • From: Joern Clausen <joern@T...>

• References:
  • Re: UTF-8+names
    • From: Mike Champion <mc@x...>

• Prev by Date: RE: UTF-8+names
• Next by Date: Re: UTF-8+names
• Previous by thread: Re: UTF-8+names
• Next by thread: Re: UTF-8+names
• Index(es):
  • Date
  • Thread