Re: About using LMP for Address translation

["Adrian Farrel" <adrian@olddog.co.uk>]

Hi Amit,

Sorry that your email got a bit lost in the background noise :-)

> Looking at the current discussion on Addressing I am bit confused about
the Addressing in GMPLS.

You are not the only one!

> According to me in out-of-band signaling, RSVP messages should be double
IP encapsulated with destination the as the data next hop TE router id.

"Double IP encapsulated"???
Do you mean RSVP in IP in IP? If that is what you mean, you are wrong.
You MAY use any encapsulation you like, but it is wrong to say SHOULD.
It is adequate to use RSVP in IP.

> Below are the steps involved in signaling:
>
> - When a PATH message is received with a ERO(containing data plane hops)
the outgoing PATH message should be double IP encapsulated.
> The top IP header destination should be TE router Id of the data next
hop.

1. See above for double encapsulation.

2. The use of the TE Router ID here is the basis for the debate.
The authors of the addressing draft are suggesting that the use of the TE
Router ID provides a well-known and stable routable control plane address
and are suggesting that its use be RECOMMENDED.
Dimitri has pointed out that *any* control plane address is adequate but
that it is sensible to use a stable routable control plane address.

> - The data next hop TE router ID can be obtained by resolving the data
hop in the ER0. This can be done by using TEDB.

It is certainly true that the TE Router ID has been advertised in close
conjunction to the TE link addresses.

> - Once the TE Router ID has been determined, the RSVP message can be
sent out (based on the routing table entry for the data next hop TE Router
ID) through the control channel.

Basically right, but I'd pick up on a few of your words. You say "based on
the routing table entry for the data next hop TE Router ID" and this may
be confusing. I think what you mean is "based on the (control plane)
routing table entry for the TE Router ID of the data next hop".

Cheers,
Adrian


> If LMP is used for resolving the translation from data plane address to
Control plane address, there has to be a 1:1 correspondance between
control & data Plane.
>
> If the translation is done based on the Routing table as I mentioned
above, there need not be any 1:1 correspondance.
>
> Plz let me know your views on this.
>
> Thanks and Regards,
> Amit.