I expect this from the doofus in $pain_in_the_butt_county but I am
surprised when I see this behavior
from large companies and I really don't understand it. Having a working
abuse/response system is beneficial
to us all including the gorillas. There is a cost to us if we're
spending expensive engineering time,
and network resources to deal with the traffic. Also there is an
intangible affect on our customers opinion
of our service.
The only thing I can think of is that they are making the decisions
about how important their abuse desk
is based solely on the cost of running that desk. They are seeing it as
a cost center and not thinking
about it's long term benefit to the entire network. I can't think of a
way to remove the incentive for this
short term thinking.
If I were the big cheese of the internet?
1. Transit providers would properly implement RFC 2827 filtering facing
their downstream single homed customers.
If you only connect to me and I send you x.x.x.0/24 down your T1 I
shouldn't be getting y.y.y.0 traffic from you.
This is easy to do.
2. Tier 1 backbone providers should be willing to de-peer non
responsive global networks. I've lost faith in
regulations to actually curb the flow but the tier 1 providers may have
the leverage to encourage good behavior.
For example if $pain_in_the_butt telco in $pain_in_the_butt country has
to start paying for transit to get to
$big_tier_1 then maybe they would clean up their act. The problem with
this is I can't think of a financial way
to get buy in to for idea from the business types in these companies.
3. There needs to be more responsible network citizenship among the
providers large enough to have an AS number.
It's harder to do ingress filtering if your customers are running BGP,
I can see reasonable cases where a
customer might throw traffic at me from source addresses that I didn't
expect. At this point you should require your customers to
police their internal network and be willing to give up on their
revenue if they refuse to do so.
Perhaps requiring a 24 hour human response to abuse@ emails as a
condition of having an AS from an RIR or as a
requirement for turning up a BGP connection? We expect a good NOC for a
peer but care less about a customer in most
cases.
4. Large eyeball networks would see the value in protecting their own
people and would implement RFC2827 as close
to their customers as possible. As soon as you can drop that packet on
the floor the better. The giant zombie
bot armies are a pain to them to.
Thats all I can think of at 4am, I bet you can see why nobody would
ever appoint me big cheese of the internet.
Sam Moats
On 2013-11-13 00:57, Hal Murray wrote:
William Herrin <b...@herrin.us> said:
That's the main problem: you can generate the report but if it's
about
some doofus in Dubai what are the odds of it doing any good?
It's much worse than that.
Several 500 pound gorillas expect you to jump through various hoops
to report
abuse. Have you tried reporting a drop box to Yahoo or Google
lately?
On top of that, many outfits big enough to own a CIDR block are
outsourcing
their mail to Google. Google has a good spam filter. It's good
enough to
reject spam reports to abuse@<hosted-by-google>
I wonder what would happen if RIRs required working abuse mailboxes.
There
are two levels of "working". The first is doesn't bounce or get
rejected
with a sensible reason. The second is actually gets acted upon.
If you were magically appointed big-shot in charge of everything, how
long
would you let an ISP host a spammer's web site or DNS server or ...?
What
about retail ISPs with zillions of zombied systems?