[OpenID] Can't we make openid.ns.sreg attribute a must in OpenID sreg responses?

[Peter Williams]

This got me thinking formally:-


> In an OpenID 1.1 request, the sreg extension is detected through the
> use
> of the "openid.sreg." prefix. In a 2.0 request, the sreg extension is
> detected by looking for a namespace alias with the correct namespace
> URI.

So, formally, there is no longer in 2.0 such a thing as the "sreg
extension" to OpenID Auth? i.e. sreg in the context of openid v2 is just
a(nother) namespace of the "AX extension" of OpenID Auth (v2.0).

Is the "sreg" AX namespace formally defined?

The sreg AX namespace is quite interesting case, as, when it was a v1
extension, the properties of the sreg class were all late bound, having
no static declaration: the class was really just a bag type. As it
morphs into an AX namespace for the v2 world, we now presumably have an
example now of a namespace declaring attributes/types using late
binding.

That is, is sreg now a kind of typelib meta-format, since it apparently
must self-describes its own types/syntaxes/constraints etc.

Or, is sreg is still really a bag of attributes that are just each just
a name-value pair, even in the AX world?