On Wed, 2 Apr 2003, Tuomas Lukka wrote:
On Wed, Apr 02, 2003 at 04:14:21PM +0200, Benja Fallenstein wrote:
Tuukka Hastrup wrote:
Could there be problems with case of the fragment identifier as well? And
should we require some canonicalization of the URIs?
No. Our job is to serialize an RDF graph canonically, so that we're
later able to unserialize it to the same graph. Canonicalizing URIs
would mean changing the graph.
It'd probably be worth making this an ISSUE in the PEG.
I'm probably not familiar enough with URI details, but this seems like an
issue to be kept in mind. Whether a program derives a URL or a user enters
it, there might be two slashes instead of one, a slash missing from the
end, "www" missing from the hostname etc.
I think libraries usually keep URIs in the string format, so changes
should happen only if the URI is derived for some reason.