Re: BGP-4 changes

[Curtis Villamizar <[e-mail]>]

In message <[e-mail]>et>, [e-mail] write
s:
> *** Resending note of 09/04/96 11:14
> Subject: RE: BGP-4 changes
> 
> >So you want to (e.g.) be able to say something like "For all MAE-E routes,
> >increment MED by 100"?
> 
> >I think that the proposed text already allows this, and the details of
> >how/when/if to do it should be left to implementors.
> 
> >But, I kinda agree with Rich that localpref and policy may be a cleaner way
> >to reflect stuff like this.
> 
> Thanks, John.
> 
> I would be worried about what the value "100" means, with respect
> to the metrics of a neighboring AS.  Does it mean two DS-3 hops
> or one hundred hops? and what if their link metrics change? etc.
> 
> You can safely drop MED attributes, but you modify MED values
> at your own risk...
> 
> -- Richard Woundy, IBM


I could see doing this if you are in agreement with whoever you are
sending MED about what the MEDs mean.  For example, some big customer
with more than one attachment might twiddle their OSPF metrics to get
traffic tuned sort of right.  If one of the provider/customer
attachments or provider links near it were running a bit hot and there
was agreement to do so, the MEDs at one attachment point might be
bumped up.  This could also apply to providers with multiple
interconnections who are willing to shuffle a bit of traffic from one
place to another to give both provider's customers better service.
This sort of cooperation is definitely not unheard of.

Curtis