User @ domain.tld as ID (Once again)

[Granqvist, Hans]

> Also, unless you can come up with a way to map an email 
> address onto a URL that doesn't involve "magic" URLs that 
> ...

How about mailto:user at domain.com? 

The reason why http://user.domain.com/ URLs work so well is 
that they specify http (the 'scheme') (even without the scheme,
such as 'user.domain.com', http is de-facto scheme today)

The 'user at domain.com' construct never reached the same standing
with common web users, since it was never really an end-user 
browser facing descriptor.

I think any identifier that is a proper URL (that is, including 
a scheme) could be valid.  Or the protocol could limit it to the 
http[s] and mailto schemes, to begin with.

Secondary abstraction of 'some URL' --> HTTP URL should be possible
but not demanded by the protocol.  Of course, this would mean protocol 
artifacts, for example cookies, cannot be specified, but must be 
abstracted into requirements, for example client-side state, that need 
to be manifested by the specific protocol implementation.

Hans