08/07/01 Internet Routing Problems
Reserved AS Numbers
This may sound odd, coming from me who is a halfway alarmist, but:
1) how is this really a 'problem'? Don't get me wrong, announcing a
reserved AS is as bad as announcing 10/8, but, what did it break? Is it
enough of a problem to spam nanog about it?
2) Did you call/email/page/carrier-pidgeon Digex/ICI about this?
1) I don't consider my post SPAM or fluf or anything else other than
operational content. A well known and respected carrier is announcing a
reserved AS. It didn't break anything here but, the announcement wouldn't
make it into anyones tables who was using 64620 in a confederation and as
a result, it very well could be breaking something somewhere. Beyond
that, it doesn't set a very good example for others. I guess I'll have to
rethink my "lower AS = higher clue" theory.
2) It's not what it broke but, what's broken that is allowing it to happen
to begin with. It's fairly apparent that someone neglected to add 64620
to a confederation statement. (And doesn't have anything in place to
prevent leaking reserved AS's either.)
3) No. I did not contact Digex/ICI directly because:
a) I'm not a direct customer.
b) I'm not a direct peer.
c) The leak has shown up in the IPMA (http://www.merit.edu/ipma/)listings
for the past 4 days and it's obvious that they're not monitoring IPMA or
don't care.
I'll worry about this once we stamp out providers that number their
router interfaces with RFC1918 space and thereby break PMTU discovery
for people who do proper ingress/egress martian filtering... 
On the other hand, we've had RFC1918 filtering wars recently - I don't
recall flame-fests about martian AS numbers lately
Valdis Kletnieks
Operating Systems Analyst
Virginia Tech
>
> This may sound odd, coming from me who is a halfway alarmist, but:
>
> 1) how is this really a 'problem'? Don't get me wrong, announcing a
> reserved AS is as bad as announcing 10/8, but, what did it break? Is it
> enough of a problem to spam nanog about it?>
> 2) Did you call/email/page/carrier-pidgeon Digex/ICI about this?1) I don't consider my post SPAM or fluf or anything else other than
operational content. A well known and respected carrier is announcing a
reserved AS. It didn't break anything here but, the announcement wouldn't
make it into anyones tables who was using 64620 in a confederation and as
a result, it very well could be breaking something somewhere. Beyond
that, it doesn't set a very good example for others. I guess I'll have to
rethink my "lower AS = higher clue" theory.2) It's not what it broke but, what's broken that is allowing it to happen
to begin with. It's fairly apparent that someone neglected to add 64620
to a confederation statement. (And doesn't have anything in place to
prevent leaking reserved AS's either.)
This has nothing to do with confederations. This was a misconfigured peer
that was a mistake anyone could have made. It was fixed and hopefully nobody
was hurt to badly by this error.
3) No. I did not contact Digex/ICI directly because:
a) I'm not a direct customer.
b) I'm not a direct peer.
c) The leak has shown up in the IPMA (http://www.merit.edu/ipma/)listings
for the past 4 days and it's obvious that they're not monitoring IPMA or
don't care.
I assure you we do care and we do try to keep up on the merit pages but
things can get hectic and things can get missed. I thank you for the
heads-up post that did in-fact lead to the problem solved. I frown upon
the technical speculation and the name bashing.
You dont? wasnt it in April? and John was talking about the same thing.
Personally, we dont see problems like this 'anymore'. We filter on any and
all reserved ASpaths. We also filter on RFC1918. And since we filter it,
we dont worry about it anymore. Personally, I applaud Verio for all the
filtering they do (and anyone else that does it).
This was also posted in April in regards to filtering peers (you can
filter your transit providers).
http://www.nielsen.net/people/christian/linx.html
http://www.irbs.com/internet/nanog/0104/0161.html
Christian