Re: Another suggestion for draft-ietf-idr-bgp4-12.txt

[Jeffrey Haas <[e-mail]>]

On Thu, Aug 23, 2001 at 01:51:01PM -0700, Enke Chen wrote:
> This "SHOULD/RECOMMENED" no longer makes sense with the introduction of
> MP_REACH and MP_UNREACH (see my original message). Is there any reason
> that we should continue this recommendation?

I would suspect that the real gist of your argument is that its easier
to serialize path attributes that have things like MP*NLRI in them
by putting the fixed things at the start of the path attributes and
then serializing as many NLRI as you can.  This does simplify your code
quite a bit.

But what happens when we add the Next Greatest Thing that we also want
to stuff as much as possible in the path attributes?

"Should" still lets us do this.

Note that the people I work with would be just as happy to relax this 
requirement, but I would rather people think about why we're relaxing it
in the first place.  I also like the idea that packets look relatively
consistant from one vendor implementation to another.

> -- Enke

-- 
Jeff Haas 
NextHop Technologies