Re: Cache control in trailers?

Am 03.02.2021 um 11:18 schrieb Poul-Henning Kamp:
> --------
> Julian Reschke writes:
>> Am 03.02.2021 um 10:59 schrieb Poul-Henning Kamp:
>>> --------
>>> Julian Reschke writes:
>>>> Am 03.02.2021 um 10:48 schrieb Poul-Henning Kamp:
>>>>> --------
>>>>> Julian Reschke writes:
>>>>>
>>>>>> 1xx messages can not have a message body.
>>>>>
>>>>> The message body does not belong to the 142, it belongs to the trailer=
>> .
>>>>
>>>> Trailers in HTTP/1.1 require chunked encoding, thus a message body. You
>>>> can't send a message body absent a final status code message.
>>>
>>> This is not "Trailers in HTTP/1.1", so that is not relevant.
>>>
>>> This is a new HTTP extension which makes it possible to transmit a respo=
>> nse with body and header parts swapped.
>>
>> I still don't get how you deploy it in HTTP/1.1.
>
> *Only* if a requestor lights whatever "bat-signal" we choose, can the responder send a 142, the chunked body and the headers, in that order.
>
> The "bat-signal" should be designed to not pass through intermediaries which do not understand its significance.  (ie: hop-by-hop header,
> mentioned in Connection:)

Well.

What you are doing here is an incompatible change of the HTTP/1.1 wire
format (yes, via opt-in). I don't think it's a good idea to deviate from
the wire format and still call the version "1.1" (or "1.x" for that matter).

Furhermore I really do not get why a change in wire format is *needed*.
As Martin said, we can mint a new final status code that indicates that
more status information will be in trailer fields.

Best regards, Julian

Received on Wednesday, 3 February 2021 10:25:32 UTC