RE: Call for Adoption: Invariants

[Spencer Dawkins at IETF <[e-mail]>]

Still with no hat ...

On Feb 6, 2018 14:48, "Praveen Balasubramanian" <[e-mail]> wrote:

+1 on “The QUIC transport draft specifically does not say anything about
which UDP ports are to be used -- QUIC is a transport after all, and there
is no reason to tie the entire transport to a port.”

Thanks to multiple people who have replied.

I understand that QUIC isn't tied to a single UDP port number. I was
thinking about the deployability story for a non-backward-compatible QUIC
version that can't be demuxed from previous versions - what I think I'm
hearing is that such a new QUIC version would require thought about getting
a new UDP port opened the way 443 seems to be today.

I'll shut up and let you guys do QUIC now :-)

Spencer



*From:* QUIC [mailto:[e-mail]] *On Behalf Of *Jana Iyengar
*Sent:* Tuesday, February 6, 2018 11:26 AM
*To:* Mike Bishop <[e-mail]>
*Cc:* Ted Hardie <[e-mail]>; Mark Nottingham <[e-mail]>;
Spencer Dawkins at IETF <[e-mail]>; Lars Eggert <
[e-mail]>; Brian Trammell (IETF) <[e-mail]>; Mikkel Fahnøe
Jørgensen <[e-mail]>; QUIC WG <[e-mail]>

*Subject:* Re: Call for Adoption: Invariants



What Mike said -- UDP/443 was allocated for HTTP over UDP, but is not
required for HTTP/QUIC. The HTTP mapping document currently says (in Section
2.1
<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fquicwg.github.io%2Fbase-drafts%2Fdraft-ietf-quic-http.html%23rfc.section.2.1&data=02%7C01%7Cpravb%40microsoft.com%7C18885d6ec0154037503208d56d977dae%7Cee3303d7fb734b0c8589bcd847f1c277%7C1%7C0%7C636535419805671598&sdata=%2B0h%2B1meATs8rcF%2F66Tb%2BvxshTFFULLfwqwoHxuy6Cb0%3D&reserved=0>)
"Servers MAY serve HTTP/QUIC on any UDP port." We can discuss this, but
let's move that discussion to where it belongs, on #495
<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fquicwg%2Fbase-drafts%2Fissues%2F495&data=02%7C01%7Cpravb%40microsoft.com%7C18885d6ec0154037503208d56d977dae%7Cee3303d7fb734b0c8589bcd847f1c277%7C1%7C0%7C636535419805681132&sdata=ams4Bh5t3rldzh4uok6YJvGlnx8o0R7ELAogQ8iIK54%3D&reserved=0>
.



The QUIC transport draft specifically does not say anything about which UDP
ports are to be used -- QUIC is a transport after all, and there is no
reason to tie the entire transport to a port.



On Tue, Feb 6, 2018 at 11:09 AM, Mike Bishop <[e-mail]> wrote:

Two notes, one against interest:  That PR is old and currently parked,
because we don’t have consensus and it’s less urgent.  And secondly,
UDP/443 is already allocated by IANA for HTTP over UDP.



*From:* Ted Hardie [mailto:[e-mail]]
*Sent:* Tuesday, February 6, 2018 10:57 AM
*To:* Spencer Dawkins at IETF <[e-mail]>
*Cc:* Mike Bishop <[e-mail]>; Brian Trammell (IETF) <
[e-mail]>; Mikkel Fahnøe Jørgensen <[e-mail]>; QUIC WG <
[e-mail]>; Mark Nottingham <[e-mail]>; Lars Eggert <
[e-mail]>
*Subject:* Re: Call for Adoption: Invariants



On Tue, Feb 6, 2018 at 6:54 AM, Spencer Dawkins at IETF <
[e-mail]> wrote:

Asking with no hat, but ...



On Mon, Feb 5, 2018 at 1:00 PM, Mike Bishop <[e-mail]> wrote:

I support adoption.  The way to change the invariants will be to mint a new
protocol, and not claim that your new protocol is a version of QUIC.  If it
happens to be startlingly similar, all well and good.



If you roll a new protocol, that's not a version of QUIC, is it obvious to
everyone but me whether you can run both protocols on UDP/443?





Mike Bishop's PR defining httpq as a scheme (https://github.com/quicwg/
base-drafts/pull/348
<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fquicwg%2Fbase-drafts%2Fpull%2F348&data=02%7C01%7Cpravb%40microsoft.com%7C18885d6ec0154037503208d56d977dae%7Cee3303d7fb734b0c8589bcd847f1c277%7C1%7C0%7C636535419805681132&sdata=wTDG5fmupXJRRSFroNyBRpcbuwjQzz5fXs%2B41F6DheA%3D&reserved=0>)
defines UDP 443 as the default for httpq.  I think that pretty much asserts
that UDP/443 is QUIC, not a forked variant.



I thought I remembered that UDP/443 was chosen because firewalls often pass
port 443, so that aids deployment. Hence, my question.





Yes.  That may cause us to consider carefully whether UDP/443 ought to be
defined for QUIC more rather than HTTP over QUIC, or even more generally.

Ted




Thanks,



Spencer