Re: [webfinger] Simplifying and unjamming WebFinger

[Mike Jones <[e-mail]>]

Tim, I’m not sure where the feeling of despair comes from, other than finishing things always seems to take longer than it should.  (I agree with you there.)

That said, I can’t support changes to WebFinger that restrict it’s applicability to only e-mail addresses, and if you’re being consistent with your desire to have OpenID Connect succeed, neither can you.  OpenID Connect uses the ability of WebFinger to return metadata about URIs that are not e-mail addresses.  See http://openid.net/specs/openid-connect-discovery-1_0.html#URLSyntax and http://openid.net/specs/openid-connect-discovery-1_0.html#host.port.example for example uses with URLs.

My take-away from this is that those of us who care about finishing WebFinger should double down in our participation to constructively help finish the IESG review process.  I would only despair if we went back to the drawing board, because that would set us back by years and destroy the chance of finishing something that’s nearly done and has been demonstrated to meet open discovery needs in practice.

                                                                -- Mike

From: [e-mail] [mailto:[e-mail]] On Behalf Of Tim Bray
Sent: Friday, June 14, 2013 10:17 AM
To: [e-mail]
Subject: [webfinger] Simplifying and unjamming WebFinger

I’ve been following the endless back and forth on this with a feeling of despair... people seem to be talking back and forth past each other.  Lots of people think there are problems, Paul thinks there are no problems.
Back in the day, when I first heard of WebFinger, it was the simplest thing imaginable: Put in an email address and get back some pointers to IDPs or whatever.  Somewhere along the way, it morphed into getting metadata about any Resource.  Which meant that you couldn’t just use an email address, you had to turn it into a URI.

And thus the problem in the spec.  You have to figure out which URI scheme to use (acct:, mailto:, device:) and this will affect the output in ways that have to be specified in Other Places.
But I don’t have a general-purpose problem about wanting metadata for arbitrary URIs.  I have several immediate pressing problems for retrieving metadata about  email addresses.   The former is now seriously getting in the way of the latter.
So, how about a WebFinger light, in which the form of the query is
/.well-known/wf-lite?email=[e-mail]<mailto:[e-mail]>
Which yields a JRD.  You could still have the rel-selection and so on. It would be easy to understand. It would be self-contained. The draft would shrink in size dramatically. It would be instantly usable by OpenID Connect.  It would probably sail through the IESG.

 -T