Fred Baker wrote:
On Nov 20, 2009, at 1:16 PM, Stig Venaas wrote:
My point is that if say the spacing is 10ms and the RTT is 100ms, you
will needlessly be trying many pairs (up to 10) before you realize that
perhaps the first you tried actually works.
My first point is that if you don't wait multiple seconds for every
timeout, you can get a reasonble RTT for the SYN/SYN-ACK and prevent
people from turning off IPv6 in the first place.
Yes, I agree with that. I only want to point out that we should not be
too agressive. Say 200ms between each might be reasonable, while I think
10ms is far too agressive.
This of course means that for people to be happy, there should not be
too many pairs to try, or we need a decent algorithm or method for
picking the most likely ones first. And of course caching. This could
maybe be done at the host/stack level or in the application.
I believe most TCP applications don't specify source addresses. Some
might store which destination address they successfully connected to
though. If the application does not bind to a specific source address,
it might be good if the stack caches which source addresses were
successful per destination (or destination prefix).
My second point is that assumptions like "the best path is through a
prefix we both use" are silly. Analysis of a common traceroute will show
that two random users tend to use two random ISPs. So a routing policy
tat starts from "assume both users are using the same prefix" is a lost
cause. Start from the assumption that you're looking for an address pair
that works, and as a second choice, works well for a given ISP pair.
I agree. Well, if you happen to have the same prefix, then it makes
sense. Also, in say an enterprise, most of the communication is probably
internal where this makes sense.
But I agree that for most communication on the Internet, it is useless.
There are some other effects though (like preferring 2001 to reach 2001,
2002 to reach 2002 etc) but there may be better ways of achieving that.
Stig
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
