W3C home > Mailing lists > Public > ietf-http-wg@w3.org > January to March 2012

Keep-Alive and 'max'

From: Martin Thomson <martin.thomson@gmail.com>
Date: Fri, 23 Mar 2012 14:57:15 -0700
Message-ID: <CABkgnnVCxBkQoKjMqvpq+XAW359CbVTq2JyZDNnjuK9UYzNtLg@mail.gmail.com>
To: HTTP Working Group <ietf-http-wg@w3.org>
Cc: Salvatore Loreto <salvatore.loreto@ericsson.com>, Greg Wilkins <gregw@intalio.com>
I'm just going through the motions of resurrection for the Keep-Alive
header [1], which might be oxymoronic.  A question arose that we
thought might benefit from some discussion.

The Keep-Alive header, as originally documented, has a 'max'
parameter.  I've even seen it produced by some implementations (Apache
did this once).  This parameter limits the use of the connection to
'n' requests in total.

Having seen this in the wild makes me a little reluctant to just
deprecate this out of hand.  However, the only conceivable use I can
think of is to prevent clients from pipelining a bazillion[2] requests
to a server.  But there are better defenses for that, namely summary
disconnection.  On the contrary, rather than being useful, I tend to
think that this is very much un-useful, and could even be discouraged.

Does anyone have a use for ``Keep-Alive: max=<n>'' ?  Or can it be
deprecated safely?

--Martin

[1] http://tools.ietf.org/html/draft-thomson-hybi-http-timeout-01
[2] spell check fact of the day: pipelining is not a word, but bazillion is.
Received on Friday, 23 March 2012 21:57:43 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:51:57 GMT