Re: #312: should there be a permanent variant of 307?

On 2011-11-03 04:45, Mark Nottingham wrote:
>
> On 03/11/2011, at 6:34 AM, Julian Reschke wrote:
>
>> On 2011-10-29 01:12, Mark Nottingham wrote:
>>>
>>> On 29/10/2011, at 2:35 AM, Yves Lafon wrote:
>>>
>>>> On Thu, 27 Oct 2011, Julian Reschke wrote:
>>>>
>>>>> <http://trac.tools.ietf.org/wg/httpbis/trac/ticket/312>
>>>>>
>>>>> So now that we have allowed UAs to rewrite a 301 POST to GET (see<http://trac.tools.ietf.org/wg/httpbis/trac/ticket/160>), the spec doesn't have a permanent redirect that always preserves the method.
>>>>>
>>>>> (We *do* have the equivalent for temporary redirects: 307).
>>>>>
>>>>> So...:
>>>>>
>>>>> 1) Is this a problem?
>>>>
>>>> First thing is... was 301 ever used to change entries in a bookmark or a link in a page? If not, then it's not a problem worth adding a new status code.
>>>
>>> +1
>>>
>>>> A 307 with a long enough cache time should be enough to redirect people.
>>>> If it is, then 2a would be the best option (in another doc)
>>>
>>> But then you have a deployment problem; it won't be backwards-compatible with most existing browsers (i.e., the redirect won't work), so it'll never get out there.
>>>
>>> AFAICT the only way to deploy would be to mint a new CC directive that means "forever" -- and we've already discussed that and decided not to go that way, IIRC.
>>>
>>> My personal .02 - I think this is close with no action, or at most a bit of prose in 307.
>>> ...
>>
>> OK, here's something we could say in the text about 307:
>>
>>       Note: this status code is similar to 302 Found, except that it
>>       does not allow rewriting the request method from POST to GET.
>>       There is no equivalent counterpart for 301 Moved Permanently.
>>       Servers that need a "permanent" variant of this status code can
>>       specify it's lifetime using Cache-Control (Section 3.2 of
>>       [Part6]), for instance by using "Cache-Control: max-age=315360000"
>>       (an expiry time ten years in the future).
>>
>
> +0 -- not against it, not sure it's really necessary (but happy to put it in if that moves us forward).
> ...

I think *some* prose around this is useful, as otherwise the same 
questions continue to come up again and again.

*If* we believe that the decorate-with-Cache-Control thingy works, we 
should say so. If we do not, we could shorten the note.

Best regards, Julian

Received on Friday, 4 November 2011 08:28:17 UTC