Hi Mark- I think I understand your first point. I think FTP is a degenerate case though, because its just like HTTP in the sense that there's basically one way that everybody knows how to use an FTP URI to get at a *document* (e.g. the FTP protocol) -- in that way, its just like HTTP. Besides, nobody has a reason to use FTP URIs (though I'm sure someone will show me why I'm wrong!)
I'm really making the argument that HTTP is special *because* the document retrieval protocol is so associated with the URI (and widely deployed, implemented, etc) - FTP is very similar in that sense but "It Just Ain't HTTP". I think the example of email addresses is more appropriate - there has been a lot of discussion about converting email addresses to descriptor documents using SMTP extensions, using HTTP conventions (along the lines what I'm proposing), etc. Even if there's a mailto: URI scheme, we don't have a defined way with SMTP to retrieve descriptor documents, so the OpenID folks have to define one if they want to use Email addresses as identifiers. (I spose others could, but that's not happening). I'd rather the complexity imposed on developers to support these other identifiers be as simple as constructing HTTP URIs. They should be free, of course, to use another way of resolving those identifiers (e.g. the SMTP extensions mentioned earlier) if that's the way the community also defines resolution to happen (in a "native" sense). I also agree with someone else's comment that the MUST wording I have is probably not appropriate. So, if someone doesn't provide an HTTP gateway definition in their specification, then so be it - makes it much harder to adopt. As for URI squatting, I'm just not seeing the connection here. Who's squatting here on what URI space? -Gabe > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Mark Baker > Sent: Friday, March 02, 2007 8:50 AM > To: Gabe Wachob > Cc: Drummond Reed; Martin Atkins; specs@openid.net > Subject: Re: Proposal for Modularizing Auth 2.0 Discovery > > Hi Gabe, > > On 2/28/07, Gabe Wachob <[EMAIL PROTECTED]> wrote: > > Basically, the Discovery Spec would specify that for any identifier > scheme > > to work with OpenID, it MUST define a way of being constructed into an > HTTP > > URI and then returning a XRDS with an HTTP GET on that HTTP URI. > > I don't understand that. Why shouldn't, say, an ftp URI be usable as > an ftp URI instead of converted to an http URI? > > My bigger concern though, is that such a mapping would be a case of > URI space squatting. > > http://esw.w3.org/topic/UriSpaceSquatting > > Mark. > -- > Mark Baker. Ottawa, Ontario, CANADA. http://www.markbaker.ca > Coactus; Web-inspired integration strategies http://www.coactus.com _______________________________________________ specs mailing list specs@openid.net http://openid.net/mailman/listinfo/specs