Ok, cool! Just bring the issue back if you see something new ;)
2015-10-21 11:13 GMT+02:00 Kewin Rausch <kewin.rausch@xxxxxxxxxxxxxx>:
You want to have a policy analogous to transmission_control_policy in the
DTP to react to ECN marked packets?
Why that?
Nope, i said i realized this while following what happens with ECN. ECN
can flag regardless of the control of the flow.
I don't want it because of ECN.
As you said, if you decide to allocate (an application/ipcp requests) a
not reliable nor controlled flow) why would you care about ECN flags?
Yeah, i know that a flow without control does not care of what happens on
the network.
I was probably confused from the existence of the "transmission_control"
policy which is present in the DTP.
Maybe I miss interpreted your idea, but at least for the use case you
propose I do not agree with it.
Yeah, as everyone which contacted my with Skype.just after i sent the mail.
No problem, guess i'll just remove it from my modifications then. :)
Cheers,
Kewin R.
On Wed, Oct 21, 2015 at 10:26 AM, Leonardo Bergesio <
leonardo.bergesio@xxxxxxxxx> wrote:
Hi Kewin,I am not sure if I got your idea 100% so let me rephrase it. You
want to have a policy analogous to transmission_control_policy in the DTP
to react to ECN marked packets?
Why that?
To me it has no sense since you already have that in DTCP (flow control).
As you said, if you decide to allocate (an application/ipcp requests) a not
reliable nor controlled flow) why would you care about ECN flags? Maybe I
miss interpreted your idea, but at least for the use case you propose I do
not agree with it. The solution is already there and use a DTP+DTCP
supported flow.
Best,
Leo
2015-10-21 10:04 GMT+02:00 Kewin Rausch <kewin.rausch@xxxxxxxxxxxxxx>:
Hi everyone,
I recently inserted a new policy hook at DTP kernel level policy set in
my branch of Irati stack implementation. Such new hook is on the receiving
'path' of the DTP, in order to have an hook not only on the transmission
side, but also on the receiving one.
Policy like the ECN marking i recently developed mark the packet
regardless of the profile used for the communication(with or without
control), but there's actually no way to realize that condition in DTP only.
True is that using only DTP does not allow to control the flow itself,
so even if you realize that there's a congestion you can't take action to
avoid it. Still this is a particular problem and there can be a more
generic case.
Realizing that something is happening on the network(and operate on the
flow) could be useful, even with the limited control offered by DTP(well,
no control).
But again, I'm thinking in the term of the specific case of ECN, to
bring an example. The idea is to have a counterpart of transmission policy
on the receiving side.
What do you think about this? :)
Cheers,
Kewin R.
--
Leonardo Bergesio
Distributed Applications and Networks Area - DANA
i2CAT Foundation - www.i2cat.net
C/ Gran Capità 2-4, office 109, Nexus I building
08034 Barcelona, Catalonia (Spain)
T: +34 93 567 9926
F: +34 93 553 2520
W: http://dana.i2cat.net
