Re: A few semantic points for HTTP/1.0 draft

John C. Mallery (JCMA@ai.mit.edu)
Wed, 30 Nov 1994 16:03:15 +0100

At 2:20 AM 30/11/94 +0100, Bob Denny wrote:
>>I was told that 503 meant server busy and that the client should continue
>>to retry after a slight delay until it could get through. Maybe I was
>>mis-informed? For now I will remove the retry behavior so that Netscape
>>fails gracefully when a 503 is encountered and not retry again. Whenever
>>someone finally figures out what to do in the 503 case I'll reimplement
>>the new solution.
>
>Lou --
>
>It was I that "created" the 503 code last summer. My intention, and the
>thrust of the discussion on www-talk, was exactly that... The server is
>temporarily too busy, try again in a "short while". I envisioned the
>"short while" to be a few seconds, not milliseconds, not minutes. So
>your response was appropriate.

The bug was that you did not get it into the spec and so you have caused
considerable
confusion when cern allocated a different number for the meltdown case.

The meltdown case is needed in two very high volume cases of important
fielded servers.

I have changed the error code returned from 502 issued by CERN early this
fall to
your PC 503. Let us not have another big round of confusion here again.

The only way we're going to avoid further confusion is with different error
codes.

So, let us stick with 503 to mean melting down don't try back soon unless I
give you
a time.

Let us add a new error code 504 (?) for use with all administrative cases,
e.g. down for maintanance
back whenever.....

Since a server that is melting down has no cycles to spare, it needs to do
minimum work to
issue the error code. For this reason, all the nice administrative frills
need to be added to
a separate error code.

We should get a decision on this and get it into the 2.0 http spec so that
2.0 compliant servers
and clients do reasonable things in these denial of service cases.

John C. Mallery
Artificial Intelligence Laboratory
Massachusetts Institute of technology
545 Technology Square, NE43-797
Camridge, MA 02139 USA

URL: http://www.ai.mit.edu/people/jcma/jcma.html