[quicwg/base-drafts] Transport parameter registry is too constraining for innovation (#3020)

[David Schinazi <[e-mail]>]

As of [draft-22](https://tools.ietf.org/html/draft-ietf-quic-transport-22#section-22.1), transport parameter identifiers are 16bit integers, with a registry where
- `[0x0000 - 0xfeff]` uses the [RFC8128 Specification Required policy](https://tools.ietf.org/html/rfc8126#section-4.6)
- `[0xff00 - 0xffff]` is reserved for experimentation

While keeping Specification Required makes sense for a good chunk of the registry space, I think there will be many experiments that are conducted before a full specification is written (some already exist today and I expect there to be many many more once the protocol is standardized).

However, those experiments only have 256 values to play with, and that is bound to cause collisions and headaches down the road.

I think the simplest solution to this problem is to use variable length integers for transport parameter identifiers. The TLS presentation language doesn't support them but since we already use VLIs for the values of several transport parameters we're no longer bound by the TLS presentation language. I think there was also support for doing this when it was discussed on #1608, but that issue was closed when the *values* were changed to VLI with #1947, without changing the identifiers.

I'm happy to write the PR if this issue gathers support.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3020