I would like to put forth a "concise rant" here.  The short story is
that *no one should use flow-in event* and we should deprecate it.

I have several reasons for saying so.
1) The flow-in semantics is hard to reason about.  We generate a
flow-in using packet-in.  And if a route cannot be found, the flow-in
is flooded?  The flooding is actually done to a packet and not a flow.
2) You do not have control over the packet in thereafter.  Some two
components can be handling the packet-in and flow-in, which refers to
the same packets actually.
3) Flow-in really does not give you much anyway.  If you want the
header fields in flow-in, just use flow.hh.  It parses those.
So, ends my quibble about flow-in.

Regards
KK


On 12 March 2010 13:56, Guanyao Huang <[email protected]> wrote:
> I think some document of authenticator module is very necessary.
> The newest version of authenticator module seems changed a lot, but
> the code for routing module remains the same.
> For example, why there are many events defined in host_event.hh, why
> there are many entries, bindings defined in authenticator.hh.
> Normally, given a mac address, or, IP address, what is the API used to
> get its datapath?
> Without such things, I feel easily get lost in the code.
> I guess many people are developing their own module based on flow_in
> event. It would be nice if they know how this flow_in event was
> generated.
>
> On Fri, Mar 12, 2010 at 1:40 PM, Srini Seetharaman
> <[email protected]> wrote:
>> Just to add to Guanyao's observations: I wanted to mention that the
>> host having two bindings is something we run across often in our
>> Stanford setup. However, we notice it only when the topology is
>> changed during runtime.
>>
>> Srini.
>>
>
> _______________________________________________
> nox-dev mailing list
> [email protected]
> http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org
>

_______________________________________________
nox-dev mailing list
[email protected]
http://noxrepo.org/mailman/listinfo/nox-dev_noxrepo.org

Reply via email to