On Tue, Dec 16, 2008 at 5:25 PM, Brian E Carpenter
<[email protected]> wrote:
>> http://bill.herrin.us/network/rrgarchitectures.html
>
> Under strategy B you say:
>> 1. Probably not compatible with UDP or TCP
>
> I think you mean "not compatible with UDP or TCP sessions
> surviving all topology changes". There's no incompatibility
> as such unless the address pair changes.

Hi Brian,

I meant the more general way. Authors of strategy B solutions should
be prepared to speak to this point. If they didn't abandon TCP and
UDP, how *did* they manage to cover the multihoming corner cases? Or
did they just abandom some of them?

Especially on the UDP side of things, Shim6 has more than a few
scenarios in which it can't re-establish the session during an address
change event. And it suffers ambiguities with protocols like FTP and
HTTP that use more than one TCP connection to complete their
communication session.

One could argue that Shim6 is at best an incomplete strategy B system.
It fails to select a legitimate B2 variant. It isn't B2a because once
the original address is given up as a locator, the GUID presented at
layer 6 is no longer globally unique.

I don't have better answers given Shim6's constraints. Rather I see
that as part of the steep price of maintaining TCP and UDP
compatibility while adding multiple address multihoming. For better
answers, I rope the DNS into the system design and let the name serve
as the GUID.


> Oh, and you could add as criticism 4:
>
> 4. Unrealistic for IPv4 due to address depletion.

In one sense that's correct. In another, it's demonstrably untrue:
anyone with two commodity broadband connections today is holding at
least one public IP address from each. Do you really expect that
product offering to disappear completely in the next two years merely
because of the address shortage? Become a by-request offering sure,
but disappear?

Let me play devil's advocate for a moment: if we add the class E
addresses to the RFC1918 pool and go with something along the lines of
B1b, we could make the 32-bit addresses last a long, long time. Those
formerly private IP addresses could be reused for each global exit
that is the first hop of the loose source route.

Strategy B has some real hidden power to it.

Regards,
Bill Herrin


-- 
William D. Herrin ................ [email protected]  [email protected]
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004
_______________________________________________
rrg mailing list
[email protected]
https://www.irtf.org/mailman/listinfo/rrg

Reply via email to