- From: Andrea Perego <andrea.perego@uninsubria.it>
- Date: Fri, 21 Dec 2007 13:36:24 +0100
- To: Jeremy Carroll <jjc@hpl.hp.com>
- CC: public-powderwg@w3.org
Hi, Jeremy. Leaving aside for the moment the main issue, I have a question on your example. If I've correctly understood, you're saying that, in mailto:jeremy.carroll@hp.com, jeremy.carroll@hp.com is the URI path. Am I right? But according to the URI specification [1], mailto: should be the URI scheme and jeremy.carroll@hp.com should be the URI authority (which can be further split into: jeremy.carroll -> URI user infos; hp.com -> URI host). Syntactically, URI (absolute) paths must always start with / . However, we do have the problem you're pointing out, but, as far as I can see, the option requiring that / *must be* excluded and the one requiring that it *must be* included are equivalent. Actually, the former has the advantage that we can consider / as implicit, and avoid repeating it any time we specify a pathStartsWith constraint. Does this makes sense? Andrea [1]http://tools.ietf.org/html/rfc3986 Phil Archer wrote: > > Hmmm... I see. > > This suggests we amend the canonicalisation steps so that missing out > the leading / becomes important (we currently say that if it's not there > in the value of pathStartsWith then the processor should add it in). > > An alternative might be to add a new property that dealt with URI > schemes where the assumptions made in the canonicalisation are not safe, > i.e. a 'Do not canonicalise' flag. Would that make you throw your hands > up in horror? I can see that it's another operational ease/semantic > formality clash. > > An alternative would be to define a set of additional properties like > includeHostNC, pathStartsWithNC etc. that would again switch off > canonicalisation. > > Incidentally, we have an outstanding comment from Thomas Roessler that > we haven't attended to yet on the issue of canonicalisation [1] so the > whole thing is up for re-trial. > > [1] http://lists.w3.org/Archives/Public/public-powderwg/2007Nov/0012.html > > > Jeremy Carroll wrote: >> >> >> I suggest that the examples and text should unify round the initial / >> being included in an includePathStartsWith >> >> I believe that in a URI like >> >> mailto:jeremy.carroll@hp.com >> >> that the path is jeremy.carroll@hp.com >> >> so that >> >> <wdr:includePathStartsWith>jeremy</wdr:includePathStartsWith> >> >> matches this URI, but, in my book (but not the current WDs) not >> >> http://example.org/jeremy >> >> for which the path is /jeremy >> >> >> jeremy (or /jeremy)
Received on Friday, 21 December 2007 12:31:48 UTC