Keepalives are already part of TCP. The complexity of adding a timer to
deal with idle connections is very small - one line of code for every
incoming segment that's probably already there to keep track of other
timeouts anyway, and one line to check that timer when the RST arrives.
I'd rather deal with this once than assume it won't be an issue - esp.
when we've already seen the behavior from middleboxes I would claim are
already misbehaving.
Joe
On 7/29/2014 11:31 PM, John-Mark Gurney wrote:
How many connections (percetage) would be protected by adding this vs.
the complexity to the protocol?
_______________________________________________
Tcpinc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tcpinc
