On 6/15/15 3:34 PM, Rob Janssen wrote:
You continuously refer to Brian's system at UCSD as "broken", and I don't like that attitude.
But it is broken! It's a fact, and while you may find it unpleasant, it's the truth. It's not a dig at Brian, UCSD or ARDC, or you.
I've asked for details on it, and have driven some discussions on how we can fix it. Tim's diagram (shim box) is one of the best ways we can deal with it with out making any changes to the gateway.
From what I understand (again there are no public docs on it), is the amprgw
software is not using kernel mode devices for any of the IPIP tunnels or the kernel routing tables. This complicates what we can do on the amprgw box in terms of tunneling traffic off of it.
Eventually I'd say we move away from the UCSD amprgw as it stands now to more of a distributed model, but I'm in favor of small steps to make routing work.
44net should be end to end, not segmented into IPIP and BGP connectivity islands. We've seen a large growth of BGP connected subnets in the last 2 years. I think it was about 11 when I got my assignment, and now we're at 53 or 54. It's clear where the growth is :)
73's