Re: [Idr] draft-spaghetti-idr-bgp-sendholdtimer - Feedback requested

[heasley <[e-mail]>]

Tue, Apr 27, 2021 at 08:22:14AM +0000, tom petch:
> It does not mean that the data plane is broken.
> If that were the case, you would get either no keepalives
> from the peer or no more acks to your own keepalives.
> <tp>
> I agree with you up to the last sentence  If the TCP window is zero, then you cannot send anything, apart from ACK and similar control packets (since ACK controls the window and so has to get through).  Your own keepalives cannot be sent, just as updates cannot be sent.  I think that this is spelt out in the I-D.

Doesn't this depend on the TCP implementation, which should send a probe
with the seq of the last sent seq (snd.nxt - 1).  unless it is an old
impl., the probe data size should be zero - ie: no seq advancement.

if it could not send KAs, these tcp sessions would have expired after ~2h.