Re: [mpls-linux-general] key for outlabel

[James R. L. <jl...@mi...>]

You are correct in your observation.  There are two motivations for the
change.  First of all the LSR MIB gives us a hint that this is the "correct"
way because the key to the outsegment table is a sequential index.  Why would
the MIB do this though?  I believe the main reason for this is the conflict
that can occur when running MPLS on an broadcast interface.  What would
it take to maintain unique keys when we receive multipla label mappings
on the same interface with the same label value, but from 2 differnt nexthops?
The straight forward answer is: outgoing interface index, outgoing label,
and nexthop ip (which could be IPv6!).    So instead of trying to hash
a 68 bit (or 84 bit!) value, why not just maintain a sequential index :-)
(BTW that is my 2nd motivation)

Jim

On Fri, Aug 30, 2002 at 06:55:03PM -0400, James Li wrote:
> Hi Jim,
> 
> I noticed in version 1.16, the radix key for outlabel is changed to a sequential index, vs the old scheme. This key is returned to the user land, and expected for a delete operation. 
> What's the motivation behind this change?
> 
> Now the user land needs to maintain a mapping from label to key. It sounds better to me, if kernel can take care of that mapping, and allow label (instead of key) as the input parametre for delete operation.
> 
> Thanks,
> 
> James
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by: OSDN - Tired of that same old
> cell phone?  Get a new here for FREE!
> https://www.inphonic.com/r.asp?r_______________________________________________
> mpls-linux-general mailing list
> mpl...@li...
> https://lists.sourceforge.net/lists/listinfo/mpls-linux-general

-- 
James R. Leu