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

Re: [good] Question about NS 1.1

• To: xml-dev@l...
• Subject: Re: [good] Question about NS 1.1
• From: "David G. Durand" <david@d...>
• Date: Fri, 12 Apr 2002 10:46:39 -0400
• In-reply-to: <200204121343.OAA01761@p...>
• References: <000501c1e213$4a272560$6501a8c0@pcukmka><3CB6E363.DAC42B0E@s...> <200204121343.OAA01761@p...>

James seems to be asking for a some kind of a static type guarantee, 
where it would be impossible to write an XSLT stylesheet that 
produced a namespace error. Or where, at least, XSLT could be asked 
to automatically ensure that its output would have all needed 
namespace declarations, and the transform writer could be completely 
unaware of the machinery of prefixes, etc.

This would be useful, but unless optional, it would make it harder 
for people to create stylesheets that produce non-namespaced, or 
non-xml results.

There are even bigger holes in the XSLT type system, such as the 
ability to emit unescaped markup -- of course those are intended to 
solve tough problems (or enable quick hacks, depending on your 
perspective), and so their use is quite clearly marked.

There is very interesting work being done on transformation languages 
that use type checkers to make guarantees about their output. This 
isn't past the research phase at this point.

Practitioners often prefer tools that let them make errors if those 
errors are the consequence of extra power. Practically, we will see 
how big a problem it is that one needs to be careful care in creating 
namespace declarations in transformations.

   -- David

At 2:43 PM +0100 4/12/02, David Carlisle wrote:
>At 3:39 PM +0200 4/12/02, james anderson wrote:
>>
>>given this expression, and given that this expression may itself be the
>>result of an arbitrary computation, i cannot predict that a transform
>>applied to the expression of a transform will produce a usable result.
>
>The result of an XSLT transform is entirely deterministic at the level
>of input tree to result tree. There is some allowed variability in the
>way the input tree can be constructed from an XML file (order of
>attributes for example) and there is some similar essentially cosmetic
>variability in the way the result may be serialised.
>
>Given that, I can't see why you should not be able to predict (or even
>specify exactly) what the result of any transform will be.

-- 
-------------------------------
David Durand                    |  12 Bassett St.
david.durand@i...        |  Providence RI, 02903-4628 USA
VP, Software Architecture       |  401-331-2014 x111 Cell: 401-935-5317
ingenta plc                     |  FAX: 401-331-2015
http://www.ingenta.com/

• Follow-Ups:
  • Re: [good] Question about NS 1.1
    • From: james anderson <james.anderson@s...>

• References:
  • RE: [good] Question about NS 1.1
    • From: "Michael Kay" <michael.h.kay@n...>
  • Re: [good] Question about NS 1.1
    • From: james anderson <james.anderson@s...>
  • Re: [good] Question about NS 1.1
    • From: David Carlisle <davidc@n...>

• Prev by Date: Re: [good] Question about NS 1.1
• Next by Date: Document Driving the Web Service
• Previous by thread: Re: [good] Question about NS 1.1
• Next by thread: Re: [good] Question about NS 1.1
• Index(es):
  • Date
  • Thread