IETF SMTP Working Group Proposal at smtpng.org

Vivien M. vivienm at dyndns.org
Wed Aug 21 23:40:08 UTC 2002


> -----Original Message-----
> From: owner-nanog at merit.edu [mailto:owner-nanog at merit.edu] On 
> Behalf Of Robert Blayzor
> Sent: August 21, 2002 7:14 PM
> To: 'Gary E. Miller'
> Cc: nanog at nanog.org
> Subject: RE: IETF SMTP Working Group Proposal at smtpng.org
> 
> 
> 
> > Uh, no.  I have seen spammers use dynamic DNS to use throw
> > away dial-ups accounts for incoming main service.
> 
> Right, but to run a "real mail server" you need a static 
> address.  Which can be registered as a valid mail server.  
> Dynamic IP's cannot.

Dynamic/static IPs, though, is a distinction that's much harder to make
these days (ahhh, how I miss the days of dialup... NOT). There are
plenty of people (myself included) who have cable/DSL connections at
home with IPs that change every 6 months or a year. Similarly, people
whose organizations can't justify the /20 from ARIN will have to
renumber their servers every time they change ISPs (how many
WorldCom/KPN Qwest/etc single-homed customers have switched or will
switch?) or outgrow the ridiculously puny allocation they were able to
justify from their upstream will have to change their "static" IPs. Oh,
and what about a DHCP setup that's set to allocate the same IP to a
certain MAC address? Is that static or dynamic? 

As for registration, well, let's try to avoid a mess like that created
by the mandatory glue record creation process involved in name server
registration, shall we? With the name server registration, you end up
having all kinds of unnecessary glue records floating around which
either a) drive someone crazy when they move their domain around, or b)
cause random people out there to end up having DNS queries showing up at
machines that aren't DNS servers (anyone care to guess how someone with
a "personal firewall" would react when they see the queies on port
53/udp?). Same thing with SWIP delegations and the like; sadly, there
are still all kinds of incorrect old information floating around in
these databases, and I'd rather not rely on some three year old
registration in deciding whether to trust some machine.

I admit that something non-IP-specific, like SSL certificates, to me
seem like a much more flexible long-term solution. Plus that way when
you renumber your mail server, you wouldn't need to reregister the new
IP, etc.

That said, I (and our several tens of thousands of users running their
own mail servers) would like to know how you define a "real mail
server". Is a "real mail server" a server that you've arbitrarily
decided needs a static IP? Is a "real mail server" a closed relay (if
so, someone on this list may feel insulted that his deliberately open
relay isn't "real" by your standards)? Is your "real mail server"
something operated by an organization with more than 200 accounts (in
which case, you're telling me that my mail server with 25 or so accounts
sitting in an Exodus colo with a perfectly static IP is not real?)? Etc.

Vivien
-- 
Vivien M.
vivienm at dyndns.org
Assistant System Administrator
Dynamic DNS Network Services
http://www.dyndns.org/ 





More information about the NANOG mailing list