Re: Timeout Status Code

John C. Mallery <JCMa@reagan.ai.mit.edu>
Errors-To: listmaster@www0.cern.ch
Date: Mon, 13 Jun 1994 19:02:38 +0200
Errors-To: listmaster@www0.cern.ch
Message-id: <19940613170108.7.JCMA@JEFFERSON.AI.MIT.EDU>
Errors-To: listmaster@www0.cern.ch
Reply-To: JCMa@reagan.ai.mit.edu
Originator: www-talk@info.cern.ch
Sender: www-talk@www0.cern.ch
Precedence: bulk
From: John C. Mallery <JCMa@reagan.ai.mit.edu>
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Re: Timeout Status Code
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas

    Date: Mon, 13 Jun 1994 10:11 EDT
    From: Tim Berners-Lee <timbl@www3.cern.ch>

    > Date: Mon, 13 Jun 1994 14:29:02 +0200
    > From: John C. Mallery <JCMa@reagan.ai.mit.edu>
    > 

    > We need a status code for internal use in clients and servers when a
    > connection times out before the transaction is complete.


    We should distinguish betweeninternal client codes and codes sent
    across the net.

Clearly.  I'm not dealing with clients at the moment.

    The HTTP status codes are sent over the net and
    have a significance only in the context of the HTTP transaction.
    So for example, if a client requests from a proxy which goes
    out to a server, and the connection between the proxy and the
    server times  out, then from the point of view of the primary
    connection between the client and the proxy, the return code
    should be something like

    500 Something went wrong: underlying service timed out

A little too general to be useful for much of anything.

    with an explanatory message including ifnromation about the
    actual error, for diagnostic purposes.  If the proxy has a code
    for "timeout", then it would need codes for all the N things
    which can happen to underlying services, from core dumps to
    ball lightning.

My immediate problem was the status code to record in the common logfile
format for my server. 500 is overly general.

A bug with numeric codes is that they do not directly encode the error
taxonomy. For examples, see the
http://clinton.ai.mit.edu/code/server/http-conditions.lisp

Are you suggesting two different timeout codes?  One for proxy servers and
another for clients (which we don't advertise)?

In any event, what does the server stuff in the log file when the client times
out?