Wendy, On 13 Feb 2013, at 14:38, Wendy Roome wrote:
> RFC 3986 does define relative URIs, so technically that's sufficient. But > I think that's like a lawyer burying some critical information -- like > "double every amount that you owe us" -- in tiny type in a footnote. > Legal, but sleazy. > > The problem is that software libraries with a simple connectToServer(uri) > function require an absolute uri. If the uri might be relative, the client > must first resolve the uri in the context of some absolute base uri. Most > libraries provide a function to do that, but it's easy to forget that step. > > Case in point: My client, which passed all the interop tests, failed when > it contacted a server that returned relative URIs. (And yes, I've since > fixed my client.) > > So if we want to allow relative URIs, we should be more explicit. At the > very least, add this sentence to the end of Ben's revised uri section: > > Relative URIs should be resolved using the URI of the Information > Resource Directory as the base URI. Or a reference to section 5 of RFC 3986 rather than re-specifying its semantics. Maybe something like: URIs can be relative and MUST be resolved according to section 5 of [RFC3986]. Ben _______________________________________________ alto mailing list alto@ietf.org https://www.ietf.org/mailman/listinfo/alto