Re: Question about H2 and chunked encoding

I've still never seen a chunk trailer yet, are they even in the wild?

Always seemed like a very bad idea to me, you can't make decisions on 
headers for policy if the header arrives at the very end.

Is there at least a limit on what headers you are prohibited from 
putting in a trailer (apart of course from framing headers) - e.g. 
Content-Type or Content-Encoding would be seriously problematic in a 
chunk trailer.

Adrien


------ Original Message ------
From: "Willy Tarreau" <w@1wt.eu>
To: "Martin Thomson" <martin.thomson@gmail.com>
Cc: "HTTP Working Group" <ietf-http-wg@w3.org>
Sent: 4/08/2017 5:07:47 PM
Subject: Re: Question about H2 and chunked encoding

>On Fri, Aug 04, 2017 at 03:01:26PM +1000, Martin Thomson wrote:
>>If you don't actually have trailers, I would send an empty DATA + ES.
>>That is (to my knowledge) perfectly interoperable.
>
>Great, that's perfect, thanks!
>
>>I'll let others speak about how well tested an empty HEADERS is, but
>>there's an important difference in the semantics if you use that.  As
>>soon as you send HEADERS with ES, you declare the presence of
>>trailers.  That they are empty is potentially confusing.
>
>That's what I fear as well. I'd rather decide to always emit an empty
>DATA+ES whenever I'd like to send an empty HEADERS if it's less risky.
>
>Thank you for the fast response!
>Willy
>

Received on Friday, 4 August 2017 05:12:38 UTC