Re: [quicwg/base-drafts] Forgery limits on packet protection (#3619)

[Nick Banks <[e-mail]>]

Another (simplier?) option could be to just update keys **much** more often. I've done performance tests with msquic on the effect of updating the key ~once per round trip. There was no noticeable effect on performance. In fact, more often than not, the key update perf tests happen to perform slightly better (within noise tolerance) than without key update.

I'm not suggesting we advocate once a round trip or anything near that; more like every 2^20 packets (or 2^30 bytes?) sent or received (including failed decryptions). As I understand it, this is an ultra conservative number as far as protecting from any kind of attacks, and should be _free_ as far as any performance impact on the connection. This also has the added benefit of exercising the key update scenario more often, which should hopefully improve interoperability of the feature.

-- 
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/3619#issuecomment-622430128