Re: [Ohai] The OHAI WG has placed draft-ohai-chunked-ohttp in state "Call For Adoption By WG Issued"

[Jana Iyengar <[e-mail]>]

I'm supportive of the work in this draft and I believe it should be adopted
by the OHAI wg.

The summary of my thoughts on the question of MASQUE v OHTTP is that this
problem is not created by chunking OHTTP, and I encourage us to engage on
the question anyway.

The decision to choose one of a MASQUE or an OHTTP relay has to be made at
design time by an app developer, and it is a significant decision. A
developer has to work across organizational boundaries to establish OHTTP
relays and gateways, and this is quite a bit of work. This is not just the
state of affairs today, this is in the premise of the use of multi-party
relays for privacy. Once these are established, not being able to extend
privacy (non-PFS, sure) for a use case that is not supported today with
OHTTP (the use cases Tommy described in his presentation) means that the
only option left is to build a completely new MASQUE client stack with a
different set of relays and potentially organizations.

I do not want to deny the use of OHTTP in cases where it is the
lower-friction and adequate privacy technology to adopt.

Fastly relays traffic today for applications that use MASQUE for efficiency
and generality, and applications that use OHTTP for simplicity. I've
engaged with developers at design time on deciding what to use. There
already is overlap between MASQUE and OHTTP applicability, and I don't
think that chunked OHTTP changes this problem. (Adding PFS to OHTTP is a
great idea that we should pursue, but it increases this overlap.)

Given the experience we are gaining as a young community still deploying
both MASQUE and OHTTP, I think this discussion however is very valuable,
and documenting the applicability of both of these technologies, in terms
of architectural, performance, and privacy tradeoffs, is useful. I
encourage this discussion, and I am happy to contribute to it. This might
even be a separate document... but either way, it shouldn't prevent chunked
OHTTP from moving forward IMO.

- jana