[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ms-ospf
Subject:    Re: Flushing LSA
From:       Dave Katz <dkatz () JUNIPER ! NET>
Date:       2001-10-19 0:06:19
[Download RAW message or body]


   Hi

   Does OSPF allows to flush LSA by just sending the header of the LSA and not
   the whole packet with LS Age set to Max Age ?

   Thanks for the help.
   Gurpreet


Coincidentally (or perhaps not) this has been a topic of much
back-room discussion lately.  The Juniper box has used this method
since day one (4+ years), on the theory that the contents of a maxaged
LSA are a no-op anyhow (and based on a rather ISIS-centric mindset by
the nameless but obvious implementor), but there is at least one
vendor that seems to have recently become allergic to this approach.

There is sufficient information present (and it is syntactically
unambiguous) for this to work, and it follows conceptually with other
places in OSPF where only an LSA header is needed (DBD, etc.)  On the
other hand, although there is no specific check in the spec for
minimum LSA length based on type, it is argued that the LSA formats
imply minimum lengths (of all of the basic LSA types) greater than 20
bytes.

It should be noted that any code that looks past the header of a
Maxage LSA is inherently broken--I've heard some claims that pulling
information from the received LSA can help clean up the route being
flushed, but you *cannot* expect the data in the Maxaged LSA to be the
same as in the previously received LSA.  It may have a higher sequence
number, or it may have the same sequence number, but different data,
and end up being "newer" due to happenstance with the LSA checksum.
Either way, you dig your own grave if you expect to get anything
useful this way.  (If you need the state, pull it out of the database,
since that's what you used to generate the route in the first place.)


--Dave

[prev in list] [next in list] [prev in thread] [next in thread]