W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1994

Re: Comments on HTTP draft [of 23 Nov 1994]

From: Ari Luotonen <luotonen@neon.mcom.com>
Date: Wed, 30 Nov 1994 10:09:11 -0800 (PST)
Message-Id: <9411301809.AA20180@neon.mcom.com>
To: frystyk@www0.cern.ch
Cc: mfc@vnet.ibm.com, fielding@avron.ICS.UCI.EDU, http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com

> > I believe proxies require Full-Requests.  Ari?

CERN proxy handles any combination of HTTP1 and HTTP0; it always
speaks back to the client using the same protocol version that the
client used, that is, up- or downgrades the protocol when necessary.
If I recall, someone reported problems with HTTP0 servers though, but
that's something in the network layer (ECONN_RESET or something like
that for remote socket when it manages to close before the proxy
writes all the headers).

> The CERN proxy talks HTTP 1.0 (almost) with proxy clients even if it
> receives a 0.9 response from the origin server. I don't know how many
> 0.9 applications are still around, but I think that at least this spec
> must have the notation documented. I think it can be taken out in futue
> releases.

I wouldn't hurry with that one.  There are a good number of visitor
counters and other gadgets that just spit out the plain file,
sometimes even *without* any reaquest, just a connection to the
correct port will trigger the response... :-).  I know that's
misbehaviour, but since it works, don't break it.  We've got better
things to do than to preach about some minor abuse like that.  Since
the code is already there it doesn't even require anybody's time to
support it anymore.

-- Cheers, Ari --
Received on Wednesday, 30 November 1994 10:10:08 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:08 EDT