open relays at Earthlink

Phil Howard phil at
Wed Aug 26 01:53:49 UTC 1998

Steve Davies wrote:

> On Fri, 21 Aug 1998, Dalvenjah FoxFire wrote:
> > One extremely simple fix that the UUnet folks appear not to have stumbled
> > upon is to firewall outgoing connections on port 25 to any hosts other
> > than a specific list of earthlink, MSN, &etc mail hosts. This would only
> > require reconfiguration on the part of the particularly obstinate customers
> > who didn't follow the directions properly in the first place, and would
> > for the most part kill off the relay hijacking that goes on from those
> > networks.

FWIW, I block port 25 on all my dialups, except to my own mail servers.
Only 2 customers complained.  One was actually a mail-only customer who
dialed another small ISP in another state.  We assisted him in changing
his configuration to using the mail server at his dialup ISP.  The other
was roaming to numerous ISPs and was a more complicated case.

> ISPs sell customers a TCP/IP connection to the Internet.  To me that means
> taking my IP datagrams and delivering them to where I address them.  I
> don't see that filtering of outbound traffic is part of such a product,
> any more than hijacking my connects to port 80 somewhere and plumbing me
> into a "transparent" web cache is.

Not all ISPs do that.  Some sell a limited service consisting of a subset
of the entire scope of possible IP packets.  Of course there are also many
that sell IP "wide open".  You can take your pick.  I elect to offer just
those services which offer what I feel to be the best combination of what
most of my customers want, and what allows me to continue to offer these
services to all customers.  Those "services" that result in my staff having
to deal with huge volumes of complaints, denial of service attacks, and
being filtered by other ISPs, I simply do not offer.

> On the other hand, I would fully support anyone's right to filter
> connections from my dialin user pool addresses if they felt that they
> needed to do that.  I would, in my personal opinion, be happy to provide
> such a person with my IP pool address ranges, or info on the domain names
> we use for that (which are easy to deduce, anyway?).

I won't need to take advantage of your offer since my mail servers are not
open for relaying.  If I or my customers receive spam from your customers,
it will either be delivered the correct way, or be delivered via a direct
connection on port 25 to a mail server that is open for relaying.  My main
goal is to block the spam using the latter method since it predominates.
But in the course of discovering all the little poorly administered mail
servers that can be used as relays, I and my customers will have to endure
tons of spam, and I will get less real work done.

> (Of course, I'd rather persuade this person than my organization deals
> responsibly with spammers - but no doubt I'd be unable to persuade some)

Most spam is sent using the "hit and run" method.  Cancelling the account
is pointless, as it probably won't be used again.  Putting a stop on the CC
number from getting further accounts may help some, but they can use other
numbers, or the numbers may be stolen, or they just go to other ISPs.

IMHO, if you want to prevent spam from being sent by your customers, that
is, if you do _not_ offer this as a service, then you need to block it.

If you do not block it, then IMHO, you are offering it (whether for pay or
for free).

> If enough people refused to take mail from my pool addresses then I guess
> my customers will be duly "encouraged" to use the provided relays. (Most
> do anyway, of course)  If only a few refuse to take the mail then most
> deliveries still work fine directly; and those few feel happy that they
> are "protected".

The majority of mail servers are run by small businesses with little to no
average technical knowledge, and rarely do much to deal with it, especially
since most spam software distributes the load rather evenly so no one server
gets hit too hard.  Getting this _large_ number of servers to clean up their
act is difficult at best due to these large numbers, and the constant arrival
of more servers.

Unlike mail servers, the greatest growth in dialup is large providers like
UUNET.  The technical knowledge

> Doesn't this arrangement make sense?

It makes sense, but it is not practical.

I have not yet added mail filtering that allows me to scan every "Received"
header for any mention of known dialup spam sources.  If/when I do, UUNET
will be one of the early ones I will have to add (depending on the growth
rate of spam).  The worst source I see this month is ATT Canada, not UUNET.

What does UUNET do to _prevent_ spam originating from people so bent on
sending it that they will disregard the policy and proceed to send spam
anyway, if the service of connecting to any port 25 is offered to them?

It's not good enough to cancel an account that has already been sacrified
by the "customer".  They don't pay and you don't get any money, but then
you don't lose anything, either.

> Regards,
> Steve Davies
> Operations, UUNET UK
> (Who is in the UUNET group but does not influence policy for UUNET US)

What we are trying to do is to influence policy, not just for UUNET anywhere,
but all others.  Influencing spammers themselves is not going to work.  So
someone else has to be influenced.  We're going to choose who to influece
based on what appears to be the most practical course to the desired end

Much talk is about network peering.  We generally don't think about it this
way, but e-mail is a form of peering, too.  Any it is getting to look like
more and more of us will have to suspend such peering in certain cases.  We
will want to influence your good paying customers to switch to whatever of
your competition discovers that they can gain these customers by applying the
kind of filtering mentioned early on, blocking port 25 access.

Phil Howard | stop3849 at stop0ads at eat27me9 at
  phil      | crash410 at suck3it6 at suck8it8 at
      at    | end8it33 at ads5suck at end9ads3 at
  ipal      | w9x2y9z0 at stop9it6 at eat17me0 at
     dot    | no7spam4 at eat56me6 at end5ads2 at
  net       | stop3710 at die9spam at eat24me3 at

More information about the NANOG mailing list