Re: Call for Closure - HTTP response version

Simon Spero <ses@tipper.oit.unc.edu> wrote
  > [regarding HTTP/1.x headers that break HTTP/1.0 clients]

  > If there are such cases, then there needs to be some emergency repairs; 
  > if there are no such cases, the following is always safe, and will always 
  > use 1.1 when available:
  > 
  > 1) clients which support HTTP/1.1 SHOULD  send 1.1 requests
  > 
  > 2) servers should echo the lesser of the request version and the 
  >    supported protocol version.
  >
  > Otherwise, I call for a coin flip. 

I agree with Simon's proposal.

Let me make some assumptions.  They may be controversial, but I haven't
seen substantial contradictory evidence:

1) The HTTP/1.1 draft is clear about which HTTP/1.1 headers cannot be
sent to HTTP/1.0 clients.

2) If an HTTP/1.1 server sends a response labeled as HTTP/1.1, but with
only HTTP/1.0-compatible headers, HTTP/1.0 clients will understand it.
(There are a few known exceptions.)

3) Based on their good-faith readings of the HTTP/1.1 spec., people in
the HTTP community are divided about what response version an HTTP/1.1
server should send in response to an HTTP/1.0 request.  This difference
alone demands clearer wording in the specification.

Since no one has described a "show-stopper" scenario, I stipulate that
the choice of response version is not a threat to interoperation.  We
are left then with a choice based on protocol aesthetics or taste.

I have stated several times my preference:  responses must be labeled
with the same version as the request.

Proponents of the other view assert that my approach will slow
deployment of HTTP/1.1.  I don't see why.  Henryk and I have both asked
for a direct response to the question, Why can't clients just send
HTTP/1.1 requests when they're HTTP/1.1 capable?  An HTTP/1.1 server
will respond in kind, and both will reap the benefits of the newer
protocol.  An HTTP/1.0 server will respond as HTTP/1.0, and both will
use that.  Having a server send an HTTP/1.1 response to a user agent or
proxy that isn't HTTP/1.1 capable does nothing that I can see to speed
HTTP/1.1 deployment or interoperation.

As the one who has launched this discussion (twice), I'm open to
compelling *technical* arguments for the other side.  If there aren't
any, I'd like someone with the opposing view to admit we're arguing
taste ("religion", if you like).  Then we can stop going around in
circles and flip a coin, as Simon suggests.

Dave Kristol

Received on Tuesday, 31 December 1996 10:30:00 UTC