Re: [quicwg/base-drafts] QPACK: Encode Largest Reference modulo MaxEntries (#1763)

[Kazuho Oku <[e-mail]>]

@dtikhonov 
> You don't need an array that large, as you can only have 4096 / 32 = 128 entries. One can do with 128-element ring buffer.

Implementation-wise that's correct. OTOH, let me note that the table size needs to be 4096 / 16 = 256 entries spec-wise so that we do not have ambiguity in if the Largest Reference refers to an entry yet to be delivered.

@afrind 
> Since one cannot really have a table full of 0 byte entries, you practically don't need even a MaxEntries/32 size ring buffer. It's a nice memory optimization to allocate say only half that many entries, and resize it when needed.

That is true. The overhead is 25% compared to the size of the header table on 64-bit architectures assuming that the ring buffer is implemented as an array of pointers. I am not sure if that is a deal blocker, but it is certainly an overhead.

-- 
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/pull/1763#issuecomment-422418166