A bit of sanity..Prefix based filtering != not caring about your
backbone. Quite the reverse. As far as trusting bgp goes, most times you
can, sometimes you can't. Prefix filtering is for the times you
can't. Everybody remembers that those times happen, right?
I remember, although it is getting to be a problem with each provider
having a 'not invented here' syndrome. I give Sprint a list of our
prefixes in one format, MCI a list of our prefixes in another format.
I'm not a big fan of centralized storage, since whois.ra.net WHOIS server
doesn't seem to be responding tonight. But it would be nice if I could
generate the information once, and let all the providers use it....
Gosh, I wonder what that could be.
If your AS is in an appropriate (your upstream providers) autnum,
and the route object for the prefix is in the <pick the registry of your
things work without a phone call. When they don't work, our noc
does a good job of helping to figure out what the problem is and getting
it fixed.. That exception is the process Jon is in.
Exactly how this works hasn't been very clear to an outsider. I've
had customers with their own AS number, who moved from one provider
to another provider. I added their ASN to our AS-MACRO in the RADB
and announced their AS and network prefixes. When their first provider
stopped announcing the AS, the customer couldn't reach ANS. A call to
the ANS NOC received a response in addition to updating the RADB we were
supposed to send ANS a seperate message telling you the AS number was
now being announced by our network (as indicated by the AS-MACRO). New
network prefixes, and new AS numbers seem to get in Ok. But moving
AS numbers still seems to be a special case. I can understand why, but
it would be nice if there was some way you could let other ISPs know
what the special cases are, so we don't get surprised.
If this is not the case, I know I'm not alone in appreciating knowing
when or not we have to send ANS a seperate note for AS number policy
On a related subject, is there any concern about cleaning obsolete
objects out of the various routing registries? Information entropy
seems out of control in most of these registries. It looks like they
could use a good dose of authority control and weeding. If you aren't
a library cataloger I suspect you've never heard of it.
Or should we just declare the registry databases a failed experiment,
and leave them as trash dumps for the archalogists to dig through.