The MPLS WG Archive

["Vijayanand C - CTD, Chennai." <vijayc@ctd.hcltech.com>]

One more point I can think of is - the retention mode . Is it OK to have
Conservative retention mode when operating in DuS Ind Req Never
configuration. Should'nt the rethention and request modes be complementary
so that this kind of situations can be avoided.


Regards,
Vijay

-----Original Message-----
From: Abhijit Gadgil [mailto:gabhijit@ee.iitb.ac.in]
Sent: Friday, January 18, 2002 10:54 AM
To: John.Brennen@marconi.com
Cc: mpls@UU.NET
Subject: Re: LDP protocol problem?


John.Brennen@marconi.com wrote :

>  Router A                          Router B
>  --------                          --------
>
>  Sends Label Mapping
>                                    Receives Label Mapping
>                                    Sends Label Release
>                                     (because Router A is not the next hop)
>  Sends Label Mapping
>   (because the hop count changed)
>                                    Router A becomes the next hop
>  Receives Label Release
>                                    Receives Label Mapping
>
>
>The problem here is that the Label Mapping message which updates the
>hop count and the Label Release message are sent more or less
>simultaneously.

Actually, the problem _may_ not arise, because if you read the note 1. At
the end of process release message (LRl.), it takes care of the scenario
when LSR A is in DU mode and B releases the mapping sent by A. In that
case, A should not send a mapping to lsr B, unless B explicitely requests
it. (Which B would indeed do, unless it is request never.) Well at the
most one can say this is inefficient behavior and not incorrect behavior.
At most what can happen is for that FEC, an LSP can never be created.

However if PHP is used with unsolicited mapping, interesting scenarios can
occur. (This occured to me quite recently)

eg.

Consider

A -- B -- C

(A, B, C tail of some LSP) And when session becomes operational between
LSR A and LSR B, but there exists no session between B and C, B will send
Implicit Null label for some FECs, which it thinks it is egressing.
However, when a session is operational with LSR C, C will egress those
FECs and will send implicit Null labels, which would result in a lot of
withdraw-release and mapping messages between B and A. Also, B has to
install correct bindings in the LIB. Is it ingeneral a good/bad idea to
use PHP when sending DU labels?

Things will not be so bad in case of DoD, because B will send Implicit
Null labels for FECs requested by A only, which _could_ be substantially
less than the ones B sends unsolicitedly.


-- 
-abhijit

Abhijit Gadgil
Graduate Student,
Dept. of Electrical Engineering,
IIT, Bombay.