Re: [Netatalk-devel] atalkd broken

[David R. <dr...@bo...>]

On Mon, Feb 18, 2002 at 04:31:32PM -0500, Joe Clarke wrote:
> On Mon, 2002-02-18 at 16:06, Sebastian Rittau wrote:
> > I know that David explained it to me, since I also asked him what that
> > patch does. Unfortunately I can't find his letter now, but I think it
> > was something on the lines of "network numbers do always start with 1 -
> > using 0 is wrong". David, could you explain it again, and maybe also
> > explains, in which way this breaks on NetBSD. (It seems to work on all
> > other platforms.)

> The range 0-65534 has special meaning.  It means that the segment should
> be auto-discovered or that no router is present.  If you start with 1,
> the special meaning is lost, and the greatest end number is 65279. 
> Therefore, 1-65534 is invalid.

As best I can tell from the original PRs for NetBSD, all recent versions
of the DDP code in the NetBSD kernel sees the 0 and thinks "Invalid range"
instead of treating the range as a special meaning. I don't know enough
about the "right thing" to know which is better.

IMHO, either we should specify STARTUP_LASTNET as htons(65279) (leaving
STARTUP_FIRSTNET as htons(1) ), or make STARTUP_FIRSTNET (and probably
STARTUP_LASTNET) a OS-dependant configure option. I'll yield to my
betters on this issue.

> This is why we should _test_ our releases before releasing them.

There's a reason why a production project has "current" and "stable". :)

David

-- 
David W. Rankin, Jr.     Christian, Husband, Father, and UNIX Sysadmin. 
   ".... As for me and my family, we will serve the LORD." Joshua 24:15