Darren Reed wrote:
> Someone pointed out that there were commands to control PVC's (pvcsif,
> pvctxctl) but what about ARP servers, etc ?
> 
> Also, how do you configure all those parts or doesn't NetBSD need that?
> 
> For example, without an ATM switch, I can get a crossover ATM setup
> working with:
> 
> clipconfig add -if ci0 -arpserver wellknown
> cliparp add -pvc 0 100 llc_routed ci0
> 
> for classical IP over ATM (FORE ATM cards in Solaris) and I get two
> interfaces:

ATM can mean different (incompatible) technologies.

1)classical IP (RFC1577)
  IETF version of IPoverATM.
  ATM is NBMA (non-broadcast, non-multicast) so that it requires
  a server for ARP (and another server for multicast) if you want to
  use SVC.
2)LANE
  ATM forum version of IPoverATM
3) simple pvc
  if we don't need fancy ATM mechanisms but are interested only in
  using ATM as a configurable point-point link, a simple PVC is
  enough. (no arp is needed.)
  the en driver supports static PVCs (ATMARP is not supported).
  the pvc interface in NetBSD extends the en driver to assign
  an individual poin-to-point sub-interface to each PVC.
  one of the advantages is that it allows to use multicast over PVC
  without any trick (this feature is essential to IPv6).

so, solaris implements (1) and NetBSD implements (3).
I think you can configure solaris not to require atmarp to
interoperate with NetBSD, but I'm not sure.

> Now, why is pvc0 a "POINTOPOINT" link ?  (well, so it is true in this case :)

To make our life easier :)  In BSD, routing and multicast are much
simpler for a point-to-point link.

> On the console, I see:
> /netbsd: atm_rtrequest: bad gateway value
> Does this all look correct ?

I believe this is a leftover debug message and you can safely ignore
it.

-Kenjiro