[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: little problem with <close-session>


Hi Juergen,

At 3:35 PM +0200 9/2/04, Juergen Schoenwaelder wrote: 
I am still not sure how serious the problem is we are trying to cure.
How many TCP connections do you have to close per second in order to
run into problems on today's operating systems? Do operators have any
insights to share whether this is a problem with existing CLI scripts?
How many TCP connections do they close per second these days?

These are all interesting questions, and I would like to know the answers (but I don't). I do know that there were real-life problems with this in the mid-1990s, but I have no more recent data.

The other reason that I cited for wanting a <close-session> command is that it removes any ambiguity about when the session has been cleanly closed on both sides, so that all resources and state can be freed. This is really a question of whether we want any concept of a session to exist at the NETCONF layer, or whether we want NETCONF to assume that the graceful close of a lower layer session indicates that the NETCONF session completed successfully.

I don't have religion about this, so if no one else sees any advantages to having a <close-session> command, please just remove it. If that is the choice of the WG, I will put a caveat in the SSH mapping about connections being left in TIME-WAIT state on the manger, and we can move on...

Margaret


--
to unsubscribe send a message to netconf-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/netconf/>