Re: RtConfig's view of router sets in inet-rtr "peer" attributes

[Mark Prior <]

At 4:13 PM +0930 22/10/02, Mark Newton wrote:
> So -- what am I doing wrong?
I don't believe you are doing something wrong per se but I think you 
are possibly doing something that no one else had tried and thus you 
have discovered that it's not implemented. Just because something is 
in the RFC doesn't guarantee that RtConfig implements it :-) Cengiz 
generally implemented thing based on now many people were whining for 
it (or how loud the whine was :-) I suspect that the RIPE NCC works 
in much the same manner.

It would seem that the code in RtConfig assumes the peer statement 
would expand to a single peer (ie it doesn't iterate through an array 
of possible peers per peer statement) but it doesn't dereference the 
value anyway so even if you only had one member in the rtr-set it 
wouldn't work.

I would suggest that you probably don't want to use RPSL to generate 
iBGP configurations anyway. Randy Bush make a statement in a RPS WG 
meeting that internal policy doesn't belong in RPSL anyway (or 
something like that, sorry Randy if I have completely cocked up the 
comment). I would suggest that you should tag the route with a 
community on ingress and then filter based on that community rather 
than generate ACLs on each router.

Mark.