Re: H2 vs responses which should not carry any payload

[Willy Tarreau <[e-mail]>]

Hi Greg,

On Sun, Oct 25, 2020 at 04:09:29PM +0100, Greg Wilkins wrote:
> Willy,
> 
> You describe having a semantic layer that can have H1 or H2 either side.
> So let's consider the case of H1->Poxy->H2.  In this case if the server
> gives you a body to a 204/304 response or a HEAD request, then you have no
> way of passing them back to the client  - as it breaks H1.  You must filter
> them in that case in your semantic layer, thus it would make sense to me to
> also filter them in the H2->Proxy>H2 case.

Yes, I totally agree on this point and that's what we want to do. However
thinking about deleting the trailers in H2->H2 worries me because I suspect
some might already use them.

In any case, it's clear that having more technical capabilities in H2 has
opened a new can of worms regarding HTTP semantics that once used to be
limited by the transport layer!

Cheers,
Willy