jyy at merit.edu
Fri Mar 18 16:36:17 UTC 1994
>I suspect that most providers have a classful network number
>within a CIDR block that does not have production hosts on it
>which they can use for this kind of testing.
I think so.
>Thus, this does not have to be a test using live traffic.
>Testing would look like:
>1. Send in specific classful routes including the "test network"
>2. Send in aggregate route
>3. Test connectivity ala Havard ...
>4. Withdraw the test network route
>5. Test connectivity ala Havard from a test station
>on the "test network"
>6. If everything is okay, then removing the rest of the specific
network number routes would seem to be okay.
I think this is a good approach and it is basically what Eric-jan
is doing. He did not do Step1 and step4 which are not necessary.
It achives the goal of detecting existing CIDR-traffic-black-holing-ASs
and moving towards the withdrawal of the living classful routes included
in this aggregate.
>It may be necessary to this on an aggregate by aggregate basis until
>all of the operational configuration issues are wrinkled out.
>This would seem to meet your safety concerns.
Yes. Again, I think this approach be used only for this stage
when we can name the ASs which are neigher do CIDR nor defaulting
yet. I still think this is a better approach than using live nets
to verify the brokeness of connectivity to CIDR routes.
More information about the NANOG