Maybe this is the most realistic solution...
> 2. Do a kind of ESMTP enhanced HTTP negotiation scheme.
Hmmm...  Similar to Dan's recent comment about HTTP/1.1.  Maybe this
would work, but would there be a consensus?  Some people seem to be
against negotiation because of efficiency issues.
The efficiency problem could be solved by designing a solution that is
only inefficient during the transition period.
But how would it work?   I just tried:
	% telnet www.w3.org 80
	...
	GET / HTTP/1.1
and it returned the HTML doc corresponding to /, instead of an error.
No good.  OK, how about:
	GET / BLURFL
Same result (got the HTML doc).  No good.  OK, how about:
	BLURFL / HTTP/1.1
Error.  Good.  But it CLOSED THE CONNECTION!  Arggh.  This would be
extremely inefficient!
> 3. having the identity and version number of the calling browser's 
> and then react in the server accordingly.
As a temporary solution, maybe.  But long term?  Yuck.
Erik