Re: 302 Redirect & POST

[I'm following up on my own reply. Sorry about that.]

Koen Holtman:
>S.N.Brodie@ecs.soton.ac.uk:
>>I am POSTing form contents (at a CERN 3.0 server) and receiving a
>>redirect.  Once the user has confirmed that it is OK to redirect,
>>should I send the POST command and the form data to the new URL,
>>or issue a GET?  I had presumed that I send the POST.
>
>Sending a POST is the correct thing to do.  The (March 8 1995) draft
>is not completely clear about this, but this seems to be the current
>consensus.

By `current consensus' above, I meant the consensus on how the (March
8 1995) draft should be interpreted.

I have just done some tests on a few browsers, and it seems that the
convention among browsers is to change the POST to a GET on redirect.
It could be argued that this behavior conforms to older internet
drafts of the http protocol (e.g. the one dated 5 Nov 1993).

So if you are a browser writer, my advice would be do to first try a
POST for the redirect, and if the POST fails, then try a GET because
of downward compatibility reasons.

If you are a CGI script writer, my advice would be to redirect a POST
request only to a script that can process both GET and POST requests.

Note that most people would agree that a mechanism to redirect a POST
on URI A to a GET on URI B would be much more useful than a mechanism
to redirect a POST on URI A to a POST on URI B.  Only the second, less
useful, mechanism is offered by the March 8 1995 draft.

Koen.