Assume TCP session between Host A and Host B.

When Host-A advertises window of 0 to Host-B, Host-B uses persist timer

to send probe packet every so many sec (in 5 to 60 sec range). This is
fine but introduces atleast 5 sec delay in B resuming the
transmissions.


Is it possible for a Host-A to send a pure window update without Host-B

probing it (as soon as A detects that its recv_win has opened). Also
assume Host-A has no data to send..can it send a pure window update
pkt?


Basically, is there a provision to overcome recovery from win=0 state
without using persist timer.


Thanks,
Vishal.

vishal.study@gmail.com wrote:
> Assume TCP session between Host A and Host B.

> When Host-A advertises window of 0 to Host-B, Host-B uses persist
> timer to send probe packet every so many sec (in 5 to 60 sec
> range). This is fine but introduces atleast 5 sec delay in B
> resuming the transmissions.


> Is it possible for a Host-A to send a pure window update without
> Host-B probing it (as soon as A detects that its recv_win has
> opened). Also assume Host-A has no data to send..can it send a pure
> window update pkt?

Absolutely. There is nothing in the TCP specs to preclude Host-A from
sending a "spontaneous" window update once a sufficient quantity of
window becomes available.

Of course, IIRC there is nothing in the TCP specs that says it must

> Basically, is there a provision to overcome recovery from win=0
> state without using persist timer.

Robustness would seem to dictate that one needs a persist timer on the
sender.

rick jones
--
a wide gulf separates "what if" from "if only"
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

Rick Jones <rick.jones2@hp.com> wrote:

> Robustness would seem to dictate that one needs a persist timer on the
> sender.

I left-out the biggest reason - the window update, as it is likley
piggy-backed on a "bare" ACK segment, is not going to be retransmitted
when that ACK is lost. Hence the requirement for a persist timer.

rick jones
--
portable adj, code that compiles under more than one compiler
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...