Re: [quicwg/base-drafts] Avoid data corruption with wrapped Largest Reference. (#2261)

[Bence Béky <[e-mail]>]

> You should really create a new issue for these things. Because this comment relates only to the issue, not the proposed fix. I haven't read the proposed changes, though I will if you think that there is a fix in there that is worth keeping after this.
> 
> Anyhow, the example is invalid here:
> 
> > another encoder implementation starts by adding 10 unique entries
> 
> Because:
> 
> > The encoder MUST NOT evict a dynamic table entry unless it has first been acknowledged by the decoder.
> 
> And you can only ever fit 5 entries in that table.

That's exactly the issue: "it has first been acknowledged by the decoder" is ambigous. "Acknowledged" is being used throughout the text with two different meanings, most often to refer to references, not insertions. The ban on evicting an entry the insertion of which has not been acknowledged, that you are quoting, only occurs once out of four different locations talking about when an entry can be evicted. The other three places clearly only ban evicting an entry with unacknowledged references (note the use of the same word), giving the impression that acknowledgement of the insertion is not necessary before eviction.

Since the design is correct if you read it right, but wording is not as clear as I think it should be, I consider this issue editorial. The change I propose in this PR is meant to be an improvement on clarity.

-- 
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/2261#issuecomment-450601057