[j-nsp] Arp publish on dynamic arp

[Richard A Steenbergen]

So, I was playing with the arp "publish" option under "family inet address 
x.x.x.x arp" when I noticed an odd behavior. The CLI seems to let you 
configure an arp "publish" without specifying an actual static mac 
address, like so:

family inet {
    address 10.0.0.1/30 {
        arp 10.0.0.2 publish;
    }
}

The implication there being that it would answer with whatever data is has 
dynamically learned for this ip address, which is actually the feature 
that I want (don't ask, just pretend that I have a sensible application 
for this :P). However, reading the documentation:

http://www.juniper.net/techpubs/software/junos/junos75/swconfig75-network-interfaces/html/interfaces-summary29.html#1014209

It says under syntax:

arp ip-address (mac | multicast-mac) mac-address <publish>;

Implying that either static "mac" or "multicast-mac" is required, and 
publish is simple an option. When testing the scenerio above, the config 
parser lets me commit it, and arp to 10.0.0.2 is still resolved correctly, 
but the local box does not answer:

07:59:46.109804  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x
07:59:48.129670  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x
07:59:50.149609  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x
07:59:52.169675  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x
07:59:54.262656  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x

As soon as I change it to:

family inet {
    address 10.0.0.1/30 {
        arp 10.0.0.2 mac 00:01:02:03:04:05 publish;
    }
}

It responds correctly:

07:54:36.877067  In VID [6: 309] arp who-has 10.0.0.2 tell x.x.x.x
07:54:36.877263 Out VID [0: 309] arp reply 10.0.0.2 is-at 0:01:02:03:04:05

So unless I'm confused this is a bug, the question is, is it:

A) A config parser bug, and publish on a dynamicly learned arp should not 
   be allowed; or
B) A bug in the arp response.

I get the feeling that it is probably A, even though I'd really like it to 
be B and be fixable. Someone want to comment one way or the other? :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)