Re: New I-D: Bootstrapping WebSockets with HTTP/3

On Fri, Aug 27, 2021 at 7:49 AM Lucas Pardue <lucaspardue.24.7@gmail.com>
wrote:

> Hey Ryan,
>
> On Fri, 27 Aug 2021, 14:49 Ryan Hamilton, <rch@google.com> wrote:
>
>> Howdy Folks,
>>
>> I've uploaded a -00 of a new draft about Bootstrapping WebSockets with
>> HTTP/3. Basically it's a trivial draft that just specifies an HTTP/3
>> version of the HTTP/2 setting from RFC 8441 (and specifies that the
>> semantics of Extended CONNECT are otherwise identical to that in RFC 8441).
>>
>>   https://datatracker.ietf.org/doc/draft-hamilton-httpbis-h3-websockets/
>>
>> This is my first time doing this process so I've quite likely done
>> something wrong. :)
>>
>
> In a limited context, I think this is mostly the correct way to do this
> and HTTP is the right WG. I do, however, think that it isn't quite so
> simple as the text makes out. There's a few minor details that need adding
> for HTTP/3, such as some text to describe how to terminate HTTP/3 streams.
>

Oh, that's a good point. At this point my brain mostly skips over such
differences between HTTP/2 and HTTP/3 but you're right that the text should
make them clear. If we think this draft is the right path forward, I'll
update it accordingly.

> In a more broad context, I wonder how (or if)  the current discussions in
> MASQUE and WebTrans about aligning on extended CONNECT over HTTP/3 factor
> in to the work here. If this piece is solely focused on WebSockets that
> might be OK, just trying to understand the pitch.
>

Yes, my particular motivation is specifically to enable WebSockets over
HTTP/3 in a server which currently supports WebSockets over HTTP/2. Since
most of the implementation is identical between both versions of HTTP
(other than the value of the SETTING) it would be trivial to support...
except that there is no spec that applies to HTTP/3 :)

I'm not terribly familiar with the discussions in MASQUE or WebTrans about
extended CONNECT. Are they intending to revise 8441 for WebSockets?

Cheers,

Ryan