> -----Original Message-----
> From: Dan Brickley [mailto:[email protected]]
> Sent: Wednesday, May 26, 2010 7:52 AM


> OK, let's look at what they have in common, if anything, from a user's
> perspective.

It seems to be yes to all these questions with some reservations. I can only 
guess for v.Next, but this is based on the charter.

> Do both allow users to be identified via http: URIs such as
> http://danbri.example.com/ ?

The Connect proposal allows users to use anything that can be resolved into an 
authority as a hint. What the user types in or clicks doesn't matter much. Only 
the identifier coming back from the provider matters and that has to be an 
https:// URI. The identifier is explicitly not meant to be a nice, easy to 
remember URI - it is an internal identifier used by the client to uniquely 
identify the user. The fact that what the user provides is nothing more than a 
hint is a significant architectural change.

I don't know how v.Next will address this, but this is a complicated matter 
which has significant usability implications.
 
> Do both allow users to be identified via email-shaped URIs?
> [email protected] (URI-ified as acct:[email protected] or similar)

Connect already does, but it is still not clear if an 'acct:' URI is needed. If 
the canonical identifier has to be an https:// URI (which is still being 
discussed), then the email address entered by the user is just a hint, which 
means it can be expressed as a 'mailto:' URI (since it will not be the subject 
of an XRD document or be used in links). But there is still a proposal for 
using 'acct:' URIs as the canonical, so this isn't settled.

> Do both allow users to based such URIs on domains they own/control, while
> allowing the heavy-lifting to be done by implementations hosted/run by
> (easily swappable - eg. homepage markup) external providers?

Connect supports this in a restricted manner. You can either operate your own 
provider, or find a provider willing to assert your own domain (which today 
doesn't exist, and something I doubt any of the large providers will offer). It 
will only work if smaller providers decide this is a good way to make money. 
The proposal removed the identifier aliasing capability in OpenID 2.0 which 
never really worked right.

I don't know what v.Next will support, but this is clearly something very 
important to those leading the effort.
 
> Is there some consistent story that can be put together for users while 'the
> marketplace' figures out the ugly under-the-cover details?

If history has anything to say, then no.

To me the core issue is the user experience when logging it, and that's still 
not solved for OpenID 2.0 after 3 years. And this was supposed to be the big 
priority for the foundation. It is not clear from the charter what v.Next will 
look like to the user. OAuth 2.0 supports 6 flows (and counting) customized to 
specific user experiences. Add to that the obvious input box and logo buttons, 
and the ability to use email addresses, and the variation in login experiences 
just gets bigger.

In addition, since OAuth is inherently an API-specific protocol, any solution 
based on that gives a huge amount of autonomy to the provider to guide their 
developers on the best user experience. The only way the foundation can get 
developers aligned is by building a recognizable brand/logo and then restrict 
how it can be used using legal means. I don't see the foundation being 
successful doing that. I do see individual companies being successful (and many 
are).

The connect proposal is optimized for two edge cases: a single provider (such 
as Facebook and Twitter, both using a flavor of the Connect proposal), and many 
providers with an "anything goes" approach - be as flexible as possible in 
accepting what the user provides. It is based on the assumption that building a 
coherent OpenID (or other) brand with logo that people see on sites and 
immediately recognize is not possible at this time, at least not by the 
communities involved.

Figuring out the 'under-the-cover' details matters very little because 
discovery (including capabilities) is now at a pretty mature point. It might 
mean a bit more work for developers (client complexity in Connect is very low), 
but it is an easy engineering challenge.

EHL




 
_______________________________________________
specs mailing list
[email protected]
http://lists.openid.net/mailman/listinfo/openid-specs

Reply via email to