Hi Ted,
The mechanism's not in TCP because the definition of a failed peer is
application dependent. A chemical plant automation application probably
has different requirements about when to treat a peer as failed than an
application that updates cute cat pictures, or one that keeps internet
routing tables up to date. The definition of peer failure is
independent of whether the peers communicate over TCP, DCCP, SCTP or
carrier pidgeon, so at some point the application's going to need to be
responsible for detecting that failure. Rather than provide a generic
mechanism that would partially solve many instances of the problem, the
TCP designers got TCP out of the way and let each application designer
do it themself.
Thanks for detailed reply!
It's now clear why the raw TCP doesn't define heartbeats.
Now let me reiterate my question: Have anyone attempted to standardise
the heartbeats since then (say as a layer on top of TCP) ?
Martin
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf