Re: [quicwg/base-drafts] Duplicates are dropped (#1624)

[ekr <[e-mail]>]

I'm with @mt here. It's not that hard to implement the now-conventional
anti-replay techniques, and if we don't do this, then we have to analyze
every single QUIC frame to guarantee it can't cause problems if replayed
(and moreover implementors need to enforce it). To take a simple example
from earlier, consider the flow control frames: if we require duplicate
suppression, then even if you miss the rule about things being
monotonically increasing, things are fine. However, if duplicates are
allowed, and you make that mistake, then attackers can just replay a flow
control frame and force you into a bad posture.

On Wed, Aug 1, 2018 at 12:23 AM, Martin Thomson <[e-mail]>
wrote:

> I'm going to hold the line on MUST. The protocol doesn't make any sense if
> you get (valid) duplicates and don't discard them. Given that you have to
> track what you are acknowledging, then the only additional cost is the
> threshold, which is nearly zero. It can even *be* zero if you use the
> minimum value from what you are tracking.
>
> What do you gain by having anything less than MUST? That is, other than
> the potential for crazy non-determinism.
>
> —
> You are receiving this because you are subscribed to this thread.
> Reply to this email directly, view it on GitHub
> <https://github.com/quicwg/base-drafts/pull/1624#issuecomment-409476533>,
> or mute the thread
> <https://github.com/notifications/unsubscribe-auth/ABD1oWWTUfo-YA7nAp5RTM_MbNfsmPxAks5uMVd6gaJpZM4Vp4qq>
> .
>


-- 
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/1624#issuecomment-409557936