🚀 go-pugleaf

James Carlson <james.d.carlson@sun.com> wrote:
> John D Loop <loop@snt.bst.bls.com> writes:
>> Thanks for your comments, James.
>> You would probably get an argument about DHCP and bridging in a WAN
>> environment.  I believe almost all the ADSL providers started off that way
>> (including BellSouth), but they got away from it as fast as they could.  DHCP
>> works well, as long as you are in a trusted environment.  Stolen IPs,
>> broadcast traffic, rogue DHCP servers, using MAC address as "authentication"
>> proved very troublesome.

> PPPoE has absolutely no security whatsoever.  It's in worse shape than
> DHCP, since at least with DHCP, there's ongoing effort to add
> security.

>>  The Redbacks (which BellSouth [and the rest of the
>> known universe] switched to) were added by the providers who would do almost
>> anything to go to a more familiar PPP implementation (!), where authentication
>> is via username/password, traffic isolated to the ppp connection, nobody can
>> steal IPs.  In short, it "looks" like dialup.

> Looks can be deceiving.  It's trivial to forge packets on PPPoE, and
> use these forged packets to highjack user sessions or snipe them out
> of existence as long as you have bridging at the Ethernet layer.

For what it's worth, I've been experimenting with PPPoE encrypted sessions
(using home-brewed extensions to the Roaring Penguin implementation) and
it works a treat.  Anyone fancy a 128-bit cypher on *all* their traffic?
With a few trivial extensions to the rfc2516 I believe it would be possible
to offer a highly-secure service, without sniffing, interception, or DoS
concerns.

cheers, josh.

josh.howlett@bris.ac.uk