OK... I've fleshed some things out a little bit better. I'll send a soft copy around. I can't make it to MN (I'm in CA and my job doesn't see a reason to send me to IETF). I've been to a couple of IETFs when they were in San Jose/San Francisco. My current thinking is simply to use the Origin AS as the routing identifier, stuffed in an "Extension Header" and leave the IPv6 addresses alone, using one address for each system. This way, we don't have to have a bunch of IPv6 addresses, and, we can accept short-term routing table growth as a tradeoff for long-term routing table shrinkage on a much larger scale. (I believe we'd be able to have more paths to more autonomous systems with less table space in the router and much smaller btrees). Imagine a router only having to know the following information in the long run: IP address of current best exit router for each NH ASN. Available AS Paths for each origin AS. Next hop address for each prefix which is originated by the local AS. BGP Attributes for each AS-Path (communities, etc.) I still need to look over HIP and the current multi6 proposals in more detail. Anyway, thanks for your feedback. If I'm missing some glaring problem, I'd like to know earlier rather than later. :-) Owen --On Wednesday, January 12, 2005 20:59 +0100 Iljitsch van Beijnum <iljitsch@muada.com> wrote:
On 12-jan-05, at 19:26, Owen DeLong wrote:
[...]
I'm thinking along the lines of a new protocol which could look up an End System Identifier against a local server and receive a response which was a list of valid Routing Tags for that destination. Sort of a cross between DNS and ARP. I don't want to ignore security and would like to see at least the option if not requirement to crypto- sign the information.
Ideally, I'd like to find a couple of people in the Bay Area so we can get together for a white-board discussion. I've got a few ideas rattling around in my head about how it might work, but, I'd like to collaborate with some others before proposing something to IETF.
The IETF multi6 wg has been looking at stuff not unlike this. If everything goes well, multi6 will be shut down, and a new wg will be formed to actually start building a protocol. This is supposed to happen at the next IETF in Minneapolis two months from now. If you've never been to an IETF meeting this might be a good time to start. :-)
However, there is a big difference between what you're saying and what's on the table at multi6: currently, the idea for multi6 and its successor is to take a bunch of regular addresses, and promote one of them to end system identifier (in your terminology). This has the advantage that you're compatible with existing TCP/IP and you can connect first and negotiate stuff afterwards. A true identifier/locator separation is harder, but would be a natural evolution of this, as it's basically nothing more than doing the same with an identifier that "happens" to be unreachable. (There are some additional complexities, of course.)
HIP (see Jeroen's message) is a very different take on a problem area that greatly overlaps, but isn't exactly the same.
If/when you have a softcopy of your ideas I'd be interested in reading it.