Vijay, Colleagues,
On 11 May 2011, at 16:53, Vijay K. Gurbani wrote:
> On 05/11/2011 10:25 AM, Richard Alimi wrote:
>> I see no mention that Location can't appear in a 5xx response:
>> http://tools.ietf.org/html/rfc2616#section-14.30
>>
>> Under the spec for the 5xx error codes (and 503 in particular), I
>> don't see anything that says one can't use a Location header:
>> http://tools.ietf.org/html/rfc2616#section-10.5
>> Indeed, for 503 it suggests that a server can use the Retry-After
>> header to indicate when it should be retried. It doesn't seem that far
>> of a leap to suggest an alternate location as well.
>
> Let's make sure; simply putting a Location header in the 503 may
> not cause browsers to follow it. Probably best to see what Ben
> or Martin say.
There's nothing to prevent HTTP 503 responses containing a Location: header.
I haven't tested it but I'd be surprised if browsers (or in fact many other
clients) followed it in practice.
> Regarding the Retry-After header, I don't know whether or not the
> HTTP user agents (browsers) honour it uniformly. In a certain sense,
> the Retry-After header leads to a sawtooth behaviour at a server,
> with the server alternating between large incoming traffic load
> and no incoming traffic load. This has been shown to cause problems
> in SIP, but I am not sure whether I can simply attribute the
> same problems to HTTP as well. We should seek the counsel of
> some HTTP folks on this, too.
IMO causing sawtooth behaviour as a result of using Retry-After is because the
server has a naive implementation. Sensible use of Retry-After would allow a
server to pace requests from clients provided the clients honoured the heading.
RFC2616 is not prescriptive on whether client need to honour Retry-After for
503s. A 503 with a Retry-After is the HTTP equivalent of a "back in 5 mins"
sign on a door, there's no guarantee that the server will be back when it says
it will and there's nothing to prevent clients from ignoring the sign and
continuing to knock.
With the exception of a browser where the user is continuously clicking the
retry button I would expect user agents to either treat 503 similar to 500 or
to honour the retry-after.
Ben
>> I would tend to think that 307 is more of an HTTP-layer concern, and
>> that we don't need to mention it in the status code table. The
>> dependency on RFC2616 should be enough to advise clients that they
>> should follow redirects.
>
> OK; sounds good.
>
> Thanks,
>
> - vijay
> --
> Vijay K. Gurbani, Bell Laboratories, Alcatel-Lucent
> 1960 Lucent Lane, Rm. 9C-533, Naperville, Illinois 60566 (USA)
> Email: vkg@{bell-labs.com,acm.org} / [email protected]
> Web: http://ect.bell-labs.com/who/vkg/
> _______________________________________________
> alto mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/alto
_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto
