Re: Context Independent URI

Williams, Stuart wrote,
> Yes, I agree, the resource denoted by a relative URI is also context
> dependent. What I am trying to pick up is that there are also some
> (syntactically) absolute URI (in that they start with a scheme name)
> that are also context dependent... eg. URI which use an unqualified
> domain name as the assigning authority; file: scheme URI which allow
> a hostname, but do not identify the namespace from which the hostname
> is taken (eg internet-domain name, DECNet, Novell IPX, Appletalk...).

Understood.

> So... is there a particular 'tweak' that you had in mind?

Well, it looks to me as tho' in the para above you've got the point over 
pretty clearly using "absolute URI", so,

  ContextIndependentURI: 
  An absolute URI SHOULD denote the same resource or concept independent
     ^^^^^^^^
  of the context(s) in which the URI is used. i.e. when used on 
  different occasions or by different users or in different locations, a
  given URI SHOULD denote the same resource or concept.

would be an improvement.

On a less pedantic note, I don't really know how to obey this SHOULD in 
any case where the authority part is a DNS name. What that resolves to 
on any host not under my control is completely determined by the local 
DNS configuration rather than my intentions in minting the URI.

Going in the other direction, this SHOULD seems to deprecate URIs which 
are only intended for consumption within, eg. an intranet with split 
horizon DNS. It seems a little harsh, and probably futile, to craft a 
SHOULD which runs against such a common and reasonable practice. You 
could try something like this,

  ContextIndependentURI: 
  An public absolute URI SHOULD denote the same resource or concept
     ^^^^^^
  independent of the context(s) in which the URI is used. i.e. when used
  on different occasions or by different users or in different
  locations, a given URI SHOULD denote the same resource or concept.

but this is problematic too, because the public/private split is 
presumably part of the context dependence you're trying to discourage.

In truth, I don't think there's any way of fixing this. The resolution 
of DNS names is by its very nature context dependent ... and that's a 
_feature_ of DNS, not a bug.

Cheers,


Miles

Received on Monday, 22 July 2002 05:58:28 UTC