Re: [Ice] Éric Vyncke's No Objection on draft-ietf-ice-pac-04: (with COMMENT)

[Barry Leiba <[e-mail]>]

>     > I also wonder why section 1 talks about 'race conditions' while the issue is
>     > related to a too short default time-out or in the use cases of section 3. (in
>     > my mind, 'race conditions' is an unusual sequence of events).

For what it's worth, a "race condition" is simply a case where results
differ based on the order that things happen (depending upon "who wins
the race").  If the results of something are different depending upon
factors such as which entity has more latency at the moment, which
process got swapped out at the wrong time, or just who happened to get
a "better" time-slice just now, we have a race condition.

In the case described in the document, if a candidate gets its "name"
in while the endpoint is still processing things, that candidate is
considered; if it's a little too late, say because of network latency
or slower processing somewhere or the like, then it doesn't get
considered.  That sounds like a race condition to me.

Barry