Subject: Re: Future of TCP/IP?
To: Rob Healey <rhealey@altair.helios.mn.org>
From: Erik E. Fair <fair@cesium.clock.org>
List: current-users
Date: 12/22/1995 14:35:31
At 7:57 12/22/95, Rob Healey wrote:

>        As classless networks become the norm the pressure on vendors
>        to provide IPv6 is lessening. Especially with alot of businesses
>        using the IETF reserved nets behind firewalls it is slowing down the
>        consumption of networks noticably.

See RFC 1627. So-called "private network numbers" are a really bad idea.
Firewalls are a temporary aberration, and proxy servers a completely evil.
I speak as one who designed, built, and operated a firewall system for a
multi-billion dollar corporation.

>        It might take alot longer than most people believe for vendors to
>        break down and provide IPv6. While this wouldn't necessarily prevent
>        NetBSD from doing it, you'd only be able to use it with other
>        NetBSD systems and would have to use IP-IP to get past core routers
>        that don't speak IPv6.

People are doing this now for interoperability testing. It is clear that
until there is a critical mass of host implementations, the router vendors
are going to sit on their hands. I believe that there are two key required
features of IPv6 that will drive its adoption, address space increase
aside:

1. totally automatic host configuration, much like a Mac does AppleTalk.
This will make renumbering networks trivially easy, and will therefore make
the compression of the core "defaultless" Internet routing tables much
easier. Of course, it will also make it much easier to switch providers...

2. IP security. Not optional. Required. Big Win. Makes evil things like SSL
go away. Probably solves the SNMPv2 security mess, too.

I would love to see both IPsec and IPv6 in NetBSD pretty much as soon as
possible. (yeah, so why don't you work on it? Well, I'm getting there...).
If nothing else, it would give the tardy host vendors a place to start.

Erik Fair