Re: [dhcwg] ARP reply vs. ARP request to announce address acquired from DHCP

> I can't think of any packet capturing facility that would not let
> you see ARP. dhcp-agent happily ARPs. In -current you can find
> the code in dhcp-arp-discovery.c

BPF wasn't passing arp-reply packets at the time.   Perhaps it does 
now, or perhaps it's handled differently in lpf.   I thought it was 
rather odd at the time, but at the time fixing it wasn't my main 
priority, so I didn't look any further.

> I can test against BSD and Solaris too but serial conning them is
> too much of a pain right now :-) Were you using a windows box to
> with your earlier test?

BSD definitely sends the ARP - I've captured it in packet traces when 
debugging.   Dunno for sure about Solaris, but it uses the BSD stack, 
so it probably does.   Linux doesn't use the BSD stack, which perhaps 
explains the discrepancy - it could be that Stevens was reporting what 
the BSD stack was already doing, rather than recommending what it 
*should* do.

So the question is, what does this tell us?   Are you saying that 
Stuart's proposal is not the right way to go, or just pointing out new 
information?   I have to say that I find Stuart's rationale for sending 
an ARP request rather than a reply fairly convincing.

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
https://www1.ietf.org/mailman/listinfo/dhcwg