Camel sighting at Google

At Building 46, off Charleston, a camel. I'm sure there's a good explanation, but for some things I just prefer the mystery.


Open Issues for Discovery / Webfinger

The problem: Discover information that joe@example.org wants to publish to the world; things like their preferred identity provider, their public avatar, public contact methods, etc. Same mechanism should basically work for joe@example.org or http://example.org/joe, no wheel reinvention.

The Webfinger session at the last IIW was quite productive in the sense that it produced a long list of open issues that need resolution. The whiteboard snapshot to the right (stitched together thanks to @factoryjoe) shows the list, albeit in low res form. Translating the notes, and giving my takes:

Starting assumption : Domain owners need to play along. We're not trying to handle the case where joe@example.org wants to be discoverable, but doesn't control example.org and the domain owner doesn't want to implement discovery.

Open Issues

Location of host-meta data: Older spec calls for this to be at /host-meta for every domain; Mark Nottingham has updated his proposal to create a /.well-known/ directory instead and put host-meta in there; I'm +10 to that.

Should discoverers try www.example.org if example.org itself doesn't support discovery? My take: No, if example.org doesn't provide the discovery info directly it can do a 3xx redirect to a site that does. Don't complicate the protocol.

Should discoverers try https: URLs first? My take: No; this is not confidential data, and if you want source verification, it's more complicated than just using SSL and there are other solutions coming down the pike that are better.

What should the protocol do with 3xx's? This clearly needs a working group convened to decide on the exact correct flavor of 3xx to use in different situations. But, don't screw over people who need to move web sites and who leave a 301 to point to a new location.

Should it support other name@domain identifiers beyond email? Yes, of course.

Proxy problems with Accept: & Vary for getting discovery data from top level domains: This goes away with /.well-known.

What should the exact template semantics be (just {id}, or {local} + {domain} be for mapping a name@domain ID to a URL? Doesn't matter, pick one.

Must the discovery data be signed to enable the pattern to work? No, clients should make their own security decisions based on the evidence given. Signing is a good idea; make it easy to accomplish.

We need to document best practices on doing all of this stuff. Yes.