From: Nathan Lane [mailto:nathan@terminus.com] Sent: Friday, September 01, 2000 3:54 PM

"Roeland M.J. Meyer" wrote:

...

The real issue is getting all those routers and switches

deployed. We

...

can then turn up the clients and servers as needed.

The need for v6, at least for me, is in the deployment of mass quantities of end nodes. That I can nearly support today with existing infrastructure. I see little compelling need to upgrade my existing routers to v6 now; I do see a need for 4000 wireless end nodes at each of our sites - those could be serviced with a few app servers in the home office that knew v6 and v4-v6 tunnels [yes, technically these would be "routers"] at the remote sites.

At six to eight weeks for each router software release for the bug scrub and six months to roll out said code, we only have time to do it twice a year. Adding a new protocol would be formiddable unless tunelling were used to virtualize the infrastructure.

1) if one is using tunneling, then we really haven't gone over to IPv6. 2) if one is using tunneling, for leaf-nodes, then how is the core going to know those leaves are using IPv6? 3) Conservatively, every recent linux node has IPv6 kernel support and every Windoze box can do IPv6 (not to mention what we have just heard from the BSD camp <g>). This is over 50% of the leaves out there (conservatively). At what point is there sufficient market penetration of the technology to consider rolling IPv4/IPv6 interoperability/capability on the core routers and switches (something short of 100%, I would hope)? 4) Is it maybe that Sun, HP, Intel, Cisco, IBM, and the telco's, aren't all quite ready yet? 5) Admittedly, I haven't had the bandwidth to follow IPv6 all that closely. But, IPv6 has been pregnent quite a long time.