From: Benjamin Black (black@layer8.net)
Date: Sat Jan 29 2000 - 11:31:02 PST
> >From my reading on IPv6, it is far from overengineered; indeed, it
> dropped a number of IPv4 features because it thought they were
> overengineered, and simplified a number of others so they were easier
> to implement.
>
And at the same time added in all manner of thorny (and unsolved) problems,
such
as source address selection. I got involved in a lengthy flamewar on the
ipng list last year which essentially revealed that the largest proponents
and designers of IPv6 were a completely seperate group from the network
operators who would be deploying it.
> I think that rolling out new network protocols is very difficult. IP
> multicast still isn't deployed at most commercial NSPs, and it's many
Numerically, that is correct. However, that is a bit misleading as a few
ISPs represent the vast majority of traffic. Sprint, UUNet, Exodus,
AboveNet, etc...all participate in MBone and use MBGP for multicast
reachability distribution. A key difference between IPv6 and multicast is
that multicast provides something people actually can use, namely efficient
streaming content delivery. IPv6 provides longer addresses which we won't
need for at least another 10 years. If you want to issues such as the
mobility support in IPv6, please see your own comments about it not being
overengineered.
I will say that one feature I *really* like about IPv6 is the reason for the
source address selection problem: the seperation of node *identity* from
node *locality*. IPv4 merges the 2 concepts which allows end stations to
source packets blindly, but allows no way to identify that multiple
addresses belong to the same host.
Ben
This archive was generated by hypermail 2b29 : Sat Jan 29 2000 - 11:34:47 PST