- From: Kazuho Oku <kazuhooku@gmail.com>
- Date: Wed, 22 Oct 2025 12:31:09 +0900
- To: Ben Schwartz <bemasc@meta.com>
- Cc: Rahul Gupta <cxres@protonmail.com>, Mark Nottingham <mnot@mnot.net>, "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>, "rory.hewitt@gmail.com" <rory.hewitt@gmail.com>
2025年10月22日(水) 10:20 Ben Schwartz <bemasc@meta.com>: > > > My concern, as Rory correctly captures, is that ultimately there are only two states being mapped on to three values. It's not wrong, but seems inefficient. > > It's possible that "?0" and (none) are identical in theory, but not in practice. That's because the meaning of (none) varies between deployments. If the deployment sometimes carries unrecognized incremental streams, then all unrecognized (none) streams must be passed incrementally today. Right. Defining Incremental: ?0 as an explicit “may buffer” signal lets intermediaries that recognize it buffer with greater confidence. This comes at minimal cost, precisely because ?0 is semantically equivalent to omitting the Incremental header; put differently, intermediaries are free to ignore Incremental: ?0. > > I agree that, in the future, it may make sense to define some parameters for "Incremental". I don't see anything strange about those parameters being limited to the "?1" case, if that turns out to be sensible*. > > --Ben > > *Perhaps a parameter like "max-interval", the max time between sends by the origin, could apply to "?0" as well. -- Kazuho Oku
Received on Wednesday, 22 October 2025 03:31:30 UTC