Re: Issue 43 (combining fragments)

tis 2009-07-28 klockan 17:36 +0200 skrev Henrik Nordstrom:
> tis 2009-07-28 klockan 17:32 +0200 skrev Henrik Nordstrom:
> 
> > And we don't allow fragments in Location or Content-Location either,
> > with both referencing absoluteURI and not URI-reference. Neither 2068,
> > 2396 or 3986 (in the form of absolute-URI) allows fragments in this
> > production.
> 
> My error. Forgot that we had changed Location to reference URI instead
> of absolute-URI.
> 
> So my question in light of what has been seen, should that change stay
> or should we revert back to absolute-URI as used by 2068 and 2616 moving
> the use of fragments back into the "undefined" field?

Reading back on the history of this thanks to archive.org restoring the
old unofficial errata list and the references from there the decision to
allow fragments in Location when used in redirects seem to be fairly
well accepted, and also that the meaning is unspecified when both the
original request URI and the Location URI has fragments.

As the server generating the redirect can not know if there was a
fragment or not in the URI accessed by the user I would even say that in
general the behavior is undefined when returning a fragment in Location.

It should also be noted that not even the HTTP/1.0 production for
Location do allow for fragments.

Regards
Henrik

Received on Tuesday, 28 July 2009 15:54:49 UTC