[Top] [All Lists]

Re: (ngtrans) Re: ipv6-smtp-requirement comments?

2001-11-11 07:01:06

*none* of the software (IP, TCP or application) supports IPv6

Funny. I thought we were discussing ``IPv6 SMTP requirements'': things
that every MTA is going to have to do as part of the IPv6 transition.

I realize a lot of people have been using the word "transition".
I personally don't think that's a very useful way to think about it, 
because I don't think that anyone who is using IPv4 is going to
make any firm plans to discontinue using IPv4 anytime soon.
I prefer the term "coexistence".

However I do think that folks are going to be begin using IPv6,
that some of them will not have IPv4 connectivity, and that folks 
using IPv6 will want to exchange mail to folks using IPv4, and 
vice versa.
Apparently you have a different goal. You want to work with existing
MTAs. You want to work with MTAs that don't know anything about IPv6.

In that case, writing a spec is rather idiotic. You have to use the
_existing_ protocols. You have to work with what the MTAs do _now_. In
other words, you have to support IPv4.

Indeed.  But you don't have to support v4 at your own MTA if you can
find a MTA that is willing to be a mail exchanger for your incoming
v4 mail and one that is willing to relay your outgoing v6 mail.

One purpose of the spec is to describe what v6-only sites need to 
do to make this work well.  Another purpose of the spec is to 
describe how new MTAs (whether v4-only, v6-only, or dual-stack)
can work optimally given a mix of v4 and v6 MTAs. 

Of course, as long as that's the situation, IPv6 is a failure. IPv6
addresses will be functionally identical to 10.* addresses.  Users with
these second-class addresses aren't actually on the Internet; they need
application-level relays to talk to the Internet. Everyone will continue
to compete for real Internet addresses: connected IPv4 addresses.

For the forseeable future, everyone will need some way to get access
to the major services that are IPv4-based: email, web, etc.  Many sites 
are already doing this with IPv4 NATs and/or application-level proxies 
so it's not difficult to conclude that this will continue to happen.
I expect that for a very long time most sites will get at least one IPv4 
address which can be used for such NATs or proxies to support legacy
applications.  However, there will also be large numbers of hosts that 
can only be directly accessed via IPv6, precisely because the availability
of large numbers of IPv6 addresses will make it possible to network those
hosts in a standard way, when due to the address shortage it is not possible 
to do this with IPv4.    And there will be applications that require use
of IPv6 because they cannot work with NATted IPv4.

It's misleading to speak of IPv6 addresses as "second-class".  IPv6 has a 
different domain of applicability than IPv4.  IPv6 makes it possible to
network a great many more devices than IPv4, and it also makes it possible
to run new kinds of applications.  I think it likely that we will see large 
numbers of IPv6-capable "appliances" - cameras, power meters, game hardware,
security systems, cell phones - anything that would benefit by remote 
control, monitoring, or access.  

IPv6 is a failure only if you expect it to replace IPv4 in a short time.
It's quite clear to all of us that this isn't going to happen.  But 
due to the address shortage and the presence of NATs, IPv4 is getting
less and less flexible and more and more difficult to use for new 
applications.  IPv6 will be successful because it lets us do things
that cannot be done with IPv4.