Re: ns5 - lexical equivalence definition

Dave Hollander wrote:
> 
> I remember explictly leaving all of the URI issues to reference to
> the URI spec and specifically that it was realistic to require implementors
> to know about URIs. 

The point is that the URI spec doesn't define lexical equivalence.  It
is reasonable to require implementors to know about the URI spec (namely
RFC 2396).  But that isn't sufficient.  It would require knowledge of
an open-ended and yet-to-be-written collection of related RFC specs.  In
fact the generic URI RFC doesn't even require that schemes define
equivalence (see section 6).  In general you can't tell from looking at
two URIs whether they refer to the same resource (consider
case-sensitivity of the filename part).

The wording in the draft is non-interoperable.  As far as I remember the
discussion in the WG, I think the sentiment was for
character-for-character equivalence (another reason to disallow relative
URIs). There was no definitive URI spec available until very recently
(August).  Without this, I don't see how an informed decision could have
been made on the issue.

This is a substantive technical issue that cannot be settled just by
saying "leave it to the relevant RFCs".

> This wording was in section 6.4 of the 0802 version
> of the spec. No significant objects raised and was retained in the while
> the rest of 6 was moved to the appendix as agreed upon by the
> editors at Montreal.
> 
> As I see it, we have two options:
> 
> 1) accept it with resolution of what URIs are by reference
> 
> 2) reject it and send the entire spec to a yet to be formed working group.
> 
> Dave
> 
> Tim writes
> >James writes
> >>I don't think the lexical equivalence definition "Note that namespace
> >>names are URIs, the governing RFCs for which contain rules for
> >>establishing lexical equivalence" is workable.  This is way too vague
> >>and open-ended for interoperability.   If some implementations treat
> >>"http://WWW.W3.ORG/" as the same as "http://www.w3.org/" and some don't,
> >>we will not have interoperability. It's not realistic to requires
> >>implementations of namespace processors to know about all URI schemes.
> >>I think lexical equivalence should just be defined as
> >>character-for-character identity.
> >
> >This is a nontrivial issue of policy.  I think that going for either
> >character-by-character equivalence (in regards which we should
> >reference the as-yet-unpublished i18n WG work) or lexical
> >equivalence per the UR* RFC's is plausible and consistent, and
> >neither seems dramatically better to me.  Who gets to make this
> >decision?  -Tim
> >

Received on Sunday, 20 September 1998 08:04:35 UTC