On Thu, 23 Sep 1999, Manuel Bouyer wrote:

> On Wed, Sep 22, 1999 at 01:02:26PM -0700, Bill Studenmund wrote:
> > While that's what netatalk does, that's not technically correct. ;-)
> > 
> > net 0 is reserved for "this net". While for any routed situation, and for
> > unrouted phase II we want any net 0 connections to go to the loopback
> > interface (which I think will then be caught by atalkd), for unrouted
> > phase 1, net 0 is the wire. :-)
> 
> As lo0 is configured with net 0, won't the kernel anserw to this for any
> interface ?

Well fixing that is another thing I'm working on. :-) I'm going to make
the loopback address have a netmask of 0xffff.ff, rather than the current
0xffff.00. Then if we're unroutered phase 2, or routed phase 1 or 2, we
add a route from 0.0/0xffff.00 -> 0.0/0xffff.ff.

> Otherwise we'll have a problem for multihomed hosts anyway, isn't it ?

??

> > Also, I think someone else made a correct comment that this really should
> > set the route to STARTUP_FIRSTNET, not 1.
> 
> Is this correct for both phase 1 and 2 ? I think the issue was still open ...

We're in phase 2 specific code at this point, if I recall correctly. :-)

In routerless phase 2, "the wire" has nets 0xff00 -> 0xfffe on it. Nodes
pick random address in that range. In routerless phase 1, net 0 is the
wire. :-)

Take care,

Bill