Re: update: http progress notification

On 27/05/2010 6:00 a.m., Henrik Nordström wrote:
> mån 2010-05-24 klockan 13:13 +1200 skrev Adrien de Croy:
>
>    
>> Tests with unmodified Firefox and IE showed no problem with multiple 102
>> responses prior to the final response as well (even though this won't be
>> an issue since the draft proposes the client explicitly advertises support).
>>      
> Hmm.. is that explicit negotiation really needed?
>    
the hop-by-hop negotiation was the justification for

a) dropping a requirement to suppress these 1xx responses for HTTP/1.0 
clients
b) not worrying about semantics of entity headers transported back in a 
1xx response.

If we drop the requirement to advertise support from the client, we 
would need to re-think these.

An alternative could be a Progress header in the request, which then 
would normally be forwarded by proxies.  Then this progress mechanism 
could work through an HTTP/1.1 proxy that didn't know about it as long 
as it forwards 1xx responses.

Regards
Adrien

> Checking.. right the specs says "Unexpected 1xx status responses MAY be
> ignored by a user agent", not MUST, so sending progress updates MAY
> break HTTP/1.1 compliant clients.
>
> The question however is if that perhaps should be a MUST?
>
> Regards
> Henrik
>
>
>

Received on Thursday, 27 May 2010 12:19:36 UTC