Re: An idea: GxSE

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Mon, 25 Jun 2001, Bill Sommerfeld wrote:

> > 	there are a lot of IPv6 implementation already out of the door.
> > 	we need to concentrate ourselves to *operational* solution that are
> > 	deployable on implementation that are available today (= NO NEW CODE).

> i do not believe that we will be able to produce a satisfactory
> multhoming solution without new code.

I don't think the geo solutions need new code. (At least mine doesn't.)

Obviously, if new code is not an option, the solution is very likely to be
less effective. Also, new code is not really a problem, as long as each site
can implement it at their own pace.

Unless I overlook something there are basically four ideas to keep the growth
of the routing table as a result from multihoming to a minimum:

1. NAT and other address rewriting ideas. Pro: works today. Con: NAT box is
   a single point of failure, sessions don't survive address change.

2. Layer 4 mobility. Pro: implemented in the hosts, support for
   multihoming at the NIC level. Con: a lot of new software.

3. Geo addressing. Pro: no (new) software necessary, works in IPv4. Con:
   routing table effect may be as little as one order of magnitude.

4. More efficient routing information encoding (such as bitmaps). Pro: same
   hardware can accomodate several orders of magnitude larger routing table.
   Con: many things: new protocol versions, new software, etc.

Obviously, a single solution can implement more than one idea, such as 1 and
2 for GxSE.

I think all of these ideas have enough potential to warrant further
investigation, even if the time scales for implementation are very
different.

Iljitsch van Beijnum