Subject: Re: -current sendmail cancer in IPv4-only kernel
To: Steve Deering <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: current-users
Date: 05/08/2000 18:48:12
In message <v04220808b53d09f5cda4@[]>Steve Deering writes
>At 4:59 PM -0700 5/8/00, Jonathan Stone wrote:

>There you go again.  It's not my uncertainty about IPv6's success that
>I am quick to mention, but rather, on those occasions when I do express
>my uncertainty, there are qualifications that I am quick to add.

Okay.  *I*, personally, have quite vivid memories of your mentioning
that IPv6 may not succeed. Quite often. I guess I recall that more
vividy than the qualifications.


>What I objected to was using my concerns that IPv6 might not succeed as
>an argument against doing something that can help it succeed (that being:
>including IPv6 in NetBSD).

Steve, you're now mis-representing me. Not for the first time in this
thread, either.

I've never said that NetBSD should not include IPv6, or not ship it.
what I *have* been vocal about is that NetBSD binary distributions
should continue to work with*OUT* IPv6: i.e., on IPv4-only kernels.
(I _have_ said NetBSD should consider shipping non-V6 kernels as well,
for people who don't want V6.  Equating that with being pro-stagnation
of the Internet is just silly).

>If people were proposing to remove support for IPv4, I could appreciate
>your unhappiness, but that doesn't appear to be the case.

Steve, the suggestion to drop support for IPv6-less kernels is
precisely what I objected to, and why I cited your frequent comment
about the uncertainty of IPv6's success. I apologize (again) for not
following up with your latest estimate of its chances.

Besides, I think what you say below says your estimate has changed
since last time I heard you give it. Fair comment?

>If the presence of IPv6 is interfering with the use of IPv4, or causing
>the system to crash, then that's a problem that should be fixed.  One would
>hope there are less radical ways to fix it than deleting IPv6 entirely
>from the system.

In the abstract, yes. But (AFAIK) the problems involved other servers
at Stanford, with older, buggier IPv6 code, which in turn causes
problems for the KAME dual-stack (or an older version of it).
The simplest solution is for me to turn off IPv6 on *my* boxes,
lest they hit the same failure mode as OpenBSD.

Note this is just me: NetBSD still ships with IPv6 on by default.
Its just that its _supposed_ to still work in IPv4-only setups.

>Why is an IPv4-only system important to you (i.e., no IPv6, no AppleTalk,
>etc.)?  Is there a technical rationale, such as memory shortage?  Or is it
>a religious thing?

Sigh. I do configure appletalk myself, at home.  I was looking for a
neutral way to say "IPv6-free".

The kernels i configure at Stanford DSG don't have either Appletalk or
IPv6.  In both casees, because there's no IPv6 or Appletalk for them
to talk to -- for reaseons which are beyond my control, which I think
you know very well.

The kernels I configure elsewhere have, till now, not included IPv6.
Mostly because *right now*, it doesn't do anything I find useful, and
I've had trustworthy reports that "early" IPv6 software breaking--
such that IPv4 connectivity suffers as well.

again, All NetBSD binary distributions and snapshots ship with IPv6
support in both kernel and userland.  The scenarios that've made me
unhappy have been where I build my own v6-free kernels, and the NetBSD
precopiled userland code doesn't deal with that as gracefully as it

>I don't doubt it.  Interestingly (to me), I have gone through the opposite
>evolution, starting off quite pessimistic, and getting more and more
>optimistic as the data comes in.  I guess the people you've been talking
>to have a different source of data, or draw their conclusions based on
>something other than data.

Interesting. Do you mean data on early deployment rates, or data on
people who've looked at IPv6 and decided not to deploy it?
How does it compare to  IPv4 growth?