PI vs PA Address Space

    Date:        Thu, 18 May 1995 09:21:02 -0700
    From:        Sean Doran <smd at cesium.clock.org>
    Message-ID:  <95May18.092114pdt.5954 at cesium.clock.org>

    I am therefore always at a loss to explain why a couple
    of fairly vocal IAB members are so stringently opposed to
    developing such technology,  [ for renumbering ]

This is unadulterated crap.

No-one on the IAB I'm aware of (which I think would include the
vocal members) has ever expressed such an opinion.

I half suspect that I am supposed to be one of those "fairly vocal
IAB members", and I know that is certainly not my view.

By all means develop renumbering technology, it would be a very
useful thing to have available, even if it wasn't required to
support PA addreses, and the issues they imply.

However, please be aware that there is a HUGE difference between

	There should not be renumbering technology
and	There is no current renumbering technology

The first is an opinion, and is what you are attributing to
some (unnamed, which makes it hard to comment) IAB members.

The second is a simple fact.

At this point we have a difference of opinion - there are some
people who believe that even if renumbering technology can be
developed for IPv4, it will never be deployed in enough places
and soon enough to make any real difference to anything, and
that its very unwise to make policy now which presumes that this
renumbering technology will somehow come into existance and
be useful.

There are others who believe that it is simply essential for this
renumbering technology to exist, or the net will cease to exist
because of the routing table size problems.

Me?  I'm in both camps, I believe both of the above.   That is,
I believe that renumbering is essential, and must come, or we
won't survive.   I also believe that its very unlikely that
renumbering for IPv4 will ever be deployed widely enough that we
can really count on using it, or expect people to be able to use it.

The solution - simple - renumbering is to be, and must be, an
integral part of IPv6, using IPv6 we can assume that everyone
has access to automatic renumbering, and we can realistically
require that they make use of it when required.

If you look at my comments on the (various) IPng lists, you'll
find that I am a truly radical proponent of renumbering, anticipating,
and demanding, automatic renumbering that goes far beyond what
most people believe possible - but even if we don't get that far,
IPv6 will certainly have renumbering that IPv4 can only dream about.

Even if we don't need IPv6 in a big hurry to solve the address
space problems, I believe we need it in a big hurry to aleviate
the routing table growth problems.

Recall that IPv6 was designed (or "chartered" may be a better word)
to solve both those problems, not just the address space problem.

kre