capabilities lookup

[Kurt Raschke]

On Nov 20, 2005, at 3:18 PM, Joaquin Miller wrote:

> Here is the proposal I promised a week or so ago.  It is timely  
> now, after Michael's discussion.
>
> We all agree on one design principle: The solution must maximize  
> the chances of rapid, widespread adoption.
>
> And we all agree that, given the other design principles (we may  
> not all agree on all of those), this means it must be extremely  
> easy for an individual to set up and maintain their own identity  
> server (whether in their own bedroom closet, at their ISP, or  
> wherever).
>
> Sadly, all four of the alternatives at http://yadis.org/wiki/ 
> Draft-002 flunk.
>
> I thought to start from what is simplest and most universal for an  
> individual who wants to set up and maintain their own identity  
> server.  I came up with: placing text on a web page.

(snip)

> The parsing of the YADIS page will be trivial.  Just scan for one  
> or more of the strings.
>
> I post this in the hope it will inspire some additional ideas.
>
> If we can get past this being 2 clunky 2 believe, and still don't  
> like it,
> then the next step is to come up with something better, ...
>
> but not too much better: something that satisfies Michael's design  
> principles.
>
> Cordially, Joaquin


I must admit, this actually makes sense.  I have seen proposal after  
proposal go by on the list, and each one, it seems, complicates  
things more and more.  Once you distill away the complexity, there's  
a basic question left: Given a URL, how do we determine what services  
it supports?  I honestly don't see where XML has to come into play  
here.  As you've shown above, it's as simple as saying "I support  
version 1 of service X and version 2 of service Y".

A while back I offered up a solution involving HTML meta and link  
tags, again striving for ease of implementation.  The counter- 
argument, of course, was that it tied YADIS to HTML.  If the solution  
to being tied to HTML is complicated XML or RDF or XRID or whatever  
else, then perhaps there need to be two forms of YADIS--one for HTML  
pages, and one for 'everything else'.  Has anyone been able to offer  
a use case for a non-HTML document?  I am not sure that it is as big  
an issue as everyone makes it out to be.

Let's say that Joe User has a LiveJournal (http://www.livejournal.com/ 
~joeuser), and a site at joeuser.com.  Now, let's also say that LJ  
implements YADIS, so that Joe can log in to a YADIS-enabled site  
using his LJ.  If he wants to also be able to authenticate with  
joeuser.com, still using LJ as his identity server, he should be able  
to do so with minor modifications to a (potentially static) page on  
joeuser.com.  Essentially, that's where we are right now with  
OpenID.  Anything more complicated is just going to hamper adoption.

-Kurt