Re: XML Guidelines -04

[Tim Bray <tbray@xxxxxxxxxxxxxx>]

Graham Klyne wrote:
> At 11:08 AM 6/4/02 -0700, Tim Bray wrote:
>
> > In particular, since protocols are going to be read by an XML 
> > processor, and since an XML processor is going to have to be able to 
> > read UTF-8 and UTF-16, the requirement to handle only one of these two 
> > actually imposes extra work 
...
> I think this is an area where XML for protocol use is different from XML 
> for documents.  I can really imagine that some protocol parsers will be 
> hand-coded. 

I understand this line of argument but I don't believe it.  Parsing XML, 
despite our best efforts, is actually kind of tricky - hard enough that 
it's a challenge even with for example the perl regexp engine.  You
have to deal with constructs like

<a
><!-- ouch --><foo:bar
 a='
 a"&lt;x>"&#27;&#x27;
' b:baz
 =
 "ble
tch:"' /></a
>

Given that the best XML processors take care of all this and fit in tens 
of K and run at I/O speeds and have voluminous public test suites and 
are free.... er, isn't it bad engineering to allow rules that will *add* 
work for implementors who do it the standard way with the standard 
known-to-be-correct tools, in order to make life easier for people who 
want to engage in what is a highly questionable engineering practice? 
On top of which, the saving in effort due to not having to handle UTF-16 
hardly seems significant (hint: convert it to UTF-8 on input). -Tim