|
Click the comments link on any story to see comments or add your own. Subscribe to this blog |
02 Mar 2011
In two previous messages we looked at the question of how hard it will be to get IPv4 address space once the original supply runs out, and how much v4 address space people really need. Today we look at e-mail and IPv6. Of all the applications on the net, mail is probably the one that is least affected by NAT, and will be the least affected by running out of v4 addresses. For one thing, mail doesn't need a whole lot of IP addresses. You can easily put 10,000 users behind mail servers on a single IP, and even a giant mail system is unlikely to need more than a few hundred IPs. (For example, all of Hotmail's inbound servers sit behind 24 IPs.) So even if you had to go buy addresses for your v4 mail servers, you wouldn't have to buy very many. Mail is by design store and forward, and does not need end-to-end connections. Most other applications expect end-to-end, which means that if you have a NAT or a firewall, you need kludges to get stuff through, which usually but not always always work. Mail forwarding is designed right in, relay or forward a message as many times as you want and it is still fine. Furthermore, current mail systems invariably separate mail submission and pickup from mail transfer. That is, when you send a message using a mail program like Outlook or Thunderbird, it submits the message to a nearby outgoing mail server (MTA, for Mail Transfer Agent), which then sends it along to the recipient's incoming MTA using the SMTP mail protocol. Once it gets there, the recipient's mail program uses POP or IMAP to pick the mail up. (In web mail, the submission and pickup happen within the web mail system's network.) On a network that used v6 internally, it would not require any special engineering to have a gateway MTA which accepted submissions and POP or IMAP pickups via IPv6 to connect to the users, and sent and received via SMTP to the rest of the world on IPv4. Today there are zillions of mail sites, all of which have IPv4 addresses, so if you want your mail to work, you have to talk IPv4. Even if there turn out to be significant networks that run only on IPv6 (something I'll believe when I see it, other than semi-isolated ones behind NATs) they'll still have IPv4 mail servers to talk to the rest of the world. Given the likely low cost of buying or borrowing IPv4 space, and the fact that an extra level of relay causes no problems at all for mail, I'd expect everyone to have v4 mail connectivity if not forever, for a very long time. What would be the incentive not to? Mail is also one of the toughest services to move to IPv6. The vast majority of attempted mail deliveries are unwanted spam or malware, so mail servers have to identify and reject as much of the unwanted mail as possible. One of the most effective ways to identify unwanted mail is IP reputation blacklists, tracking IP addresses that are unlikely to send wanted mail. For hosts with sufficiently poor reputations, a mail host can reject attempted mail deliveries, without going through the relatively expensive process of receiving and filtering the messages. Using well run sources of IP reputation data such as the Spamhaus lists, a mail server can reject 90% of attempted deliveries. Since receiving and filtering mail is by far the most expensive part of mail handling, these rejections mean a close to 90% decrease in the cost of running a mail system compared to receiving and filtering everything. At this point, nobody knows how to do IPv6 reputation. Part of the reason is that we have no idea how people will use their v6 address space for mail. A single address per host, as in IPv4, is only one possibility. In the common situation that a host is allocated a 64 bit address range, it could use a different IPv6 address for every message it ever sent. Spammers will surely do that, and legitimate list managers might also be tempted, to improve tracking and bounce management. IP reputation data has for over a decade been published in the DNS. (I finally wrote an RFC defining the syntax last year.) But a key reason those DNS lookups work is that mail comes from a relatively small set of addresses, so normal DNS cache behavior keeps the total DNS traffic to a tolerable level, handling repeated lookups for the same IP without going back to the master DNS server. If every spam has a different address, and requires a different DNS lookup, the increased load will be too much for most DNS caches. Most mail servers use other DNS techniques to check incoming traffic, such as a reverse DNS lookup to find the names of connecting hosts, which have the same cache busting problem. We will eventually figure out both how people use IPv6 addresses for mail, and how to manage and publish v6 reputation data (I've been doing some experiments, which I'll blog about when I have enough results), but until then, running a mail server on v6 will be a lot harder than running one on v4. And since you'll be able to handle all the real mail on v4, why bother?
|
TopicsMy other sitesOther blogsCAUCE A keen grasp of the obvious Related sitesCoalition Against Unsolicited Commercial E-mail |
© 2005-2024 John R. Levine.
CAN SPAM address harvesting notice: the operator of this website will
not give, sell, or otherwise transfer addresses maintained by this
website to any other party for the purposes of initiating, or enabling
others to initiate, electronic mail messages.