The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2002-Sep> msg00067



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Clarification on Handling a specific event in LDP

  • From: Eric Rosen <erosen@cisco.com>
  • Date: Wed, 18 Sep 2002 13:42:11 -0400
  • cc: "MPLS (E-mail)" <mpls@UU.NET>
  • User-Agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.3(Unebigoryōmae) APEL/10.3 Emacs/21.2(sparc-sun-solaris2.8) MULE/5.0 (SAKAKI)

Jack> It's an implementation detail for  any implementor whose LSR gives out
Jack> NULL labels 

No, it's only  an issue if the  LSR gives out NULL labels  on sessions which
are  running in  ordered DoD  mode.   If there  aren't any  such LSRs,  then
there's no problem in practice.  (Even  if the are such LSRs, it's still not
clear that there's much of a problem in practice.)

Jack> (1) Avoid the issue -- don't send NULL labels.  :-)

That works, but may have some costs.

Jack> (2) Leave the NULL label in place; take no action. 

This doesn't work, as Dan points out.

Jack> (3)  Withdraw the NULL label; readvertise using a non-NULL label
Jack>      only if the session supports Unsolicited advertisement. 

This isn't the case we're talking about.

Jack> (4)  Withdraw the NULL label; readvertise using a non-NULL label
Jack>      regardless of the session's advertisement setting. 

This will probably upset the peer.

Jack> (5)  Advertise using a non-NULL label, without withdrawing the
Jack>      NULL label.  The upstream peer *might* realize what you are
Jack>      trying to do...

This will probably upset the peer. 

Once you withdraw a label, you have to wait for the upstream peer to request
a new  label.  The upstream peer  could probably get away  with requesting a
new  label immediately,  not percolating  the withdraw  upstream  unless the
new  request is  refused.  Or the  upstream  peer could  just percolate  the
withdraws upstream,  in which  case the ingress  would have to  reinitiate a
request.  I  don't  really see  what  else  could  be done  without  causing
interoperability problems.