Re: Namespace names: a modification of a semi-serious proposal

> For those people who want both identification *and* retrival; 
> they can use "http:"   For those other individuals who want
> just identification, we can use "data:" and avoid the extra 
> costs.  This fits nicely with the you-pay-for-what-you-use
> phlisophy.  It also does not distort the behavior people 
> expect when they see an "http:\\" uniform resource *locator*

No, I'm sorry but it entirely misses the point about why URI
were chosen to name namesapces in the first place.

The requirement was for authors to allocate themselves a unique
name without the use of a central name registry.

data: does not provide that unless you specify what the form of the
data is (eg an fpi, or a some other proposed mechanism for generating
names) 

Specifying namespace equivalence by comparing the bytes returned by
an http url is just horribly expensive. Apart from the cost
(compossibility for some systems) of fetching the thing, what some
people want to put at the namespace URI is an XML schema. Schema are
going to be very useful but they are also going to be very large
it would be reaaly really strange if XML which was in many ways
motivated by the desire to be able to parse a document instance
without the cost of retrieving and parsing a DTD grows into
XML+Namespaces in a form that to parse a document you have
a comparable cost of retreiving and comparing schema or other
documents located at some URL.


David

Received on Sunday, 28 May 2000 07:44:14 UTC