The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Jul> msg00062



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

draft-ietf-mpls-te-feed-00.txt

  • From: "Peter Ashwood-Smith" <petera@nortelnetworks.com>
  • Date: Mon, 10 Jul 2000 09:15:51 -0500
  • X-Orig: <petera@americasm01.nt.com>

Title: RE: draft-ietf-mpls-te-feed-00.txt

   Alex,

   Thanks for the comments. The thing to keep in mind here is that how an implementation chooses to store the feedback information is not really part of the draft, and does not need to be standardized. The important point is that you receive feedback and you use that feedback information in subsequent computations until you get another LSA. There are a variety of different ways to accomplish this but probably the easiest is to store feedback information separately from flooded information. In any case, fed back information is never re-flooded, only the originator can flood its bandwidth information.

    If however you would prefer that I add a sentence or two  to make this more clear I will be happy to do so if you want to send me something.

    Cheers,

    Peter

    -----Original Message-----
    From:   Alex Zinin [SMTP:azinin@cisco.com]
    Sent:   Monday, July 10, 2000 1:59 AM
    To:     mpls
    Subject:        draft-ietf-mpls-te-feed-00.txt


      Peter, et al.

      In your draft you recommend to change the contents
      of IGP's LSDB whenever the feedback from the signaling
      component is received. I'm afraid it is not quite good.

      1. If the checksum of the LSA/LSP is not recalculated
         after the contents were changed, the IGP may report
         a fatal error if the checksum is verified before
         the next version of the LSA/LSP arrives.

      2. Recalculating the checksum is also wrong, because:

          a) only the originating router must do this

          b) if this is done by other routers it may affect
             LSA/LSP version comparison algo.

      Because of the above and because it is a 'bad thing' to
      modify link-state PDUs originated by others, I think it
      is safer and makes some more sense to say that implementations
      must account for a method to link the information provided
      by the signaling component to the corresponding elements
      of the TE database, instead of overwriting it.

      Cheers,
     
    --
    Alex Zinin