[c-nsp] ME3600 BGP Route-Maps and IPv6 (WAS: Re: preference on bgp route advertisements)

[Phil Mayers]

On 03/08/2012 07:44 AM, Gert Doering wrote:

>> If I have a route-map that parses IPv6 routes, but does not match any
>> IPv6 routes (no match ipv6 ... defined anywhere in any of the route-map
>> sequence entries) then it matches on the first _IPv4_ route map entry
>> and sets the community of that IPv6 route to the IPv4 match instead.
>> That's the bug :)
>
> As far as I have interpreted this behaviour: for an IPv6 route, the
> "match ip" statements are just not evaluated, as if "not there at all",
> and vice versa for IPv4 routes and "match ipv6".

Yeah, this behaviour is pretty well documented, and I found it quite 
surprising the first time I ran into it:

http://www.cisco.com/en/US/tech/tk365/technologies_tech_note09186a008047915d.shtml#cmdsredist

The occasion I ran into it was an attempt at laziness, to use the same 
route-map for "redis connected" and "redis static". I wanted to use a 
tag on static routes to signal "no-export" and wrote a route-map like this:

route-map redis2bgp permit 10
   match tag 100
   set community no-export
route-map redis2bgp ...

Of course, this fails for "connected" routes; because "match tag" is not 
a "supported command" for connected, it's just ignored, meaning the 1st 
statement matches for all connected routes.

Basically - match statements that are inapplicable are just IGNORED as 
opposed to the match FAILING.