News reports say that the Israeli government is close to passing a law that requires portable e-mail addresses, similar to portable phone numbers. Number portability has been a success, making it much easier to switch from one provider to another, and address portability might ease switching among ISPs. But e-mail is not phone calls. Is it even possible?

The bill’s sponsors apparently assume that e-mail messages work enough like phone calls that whatever they do to make phone numbers portable can work the same way for mail. Unfortunately, they’re wrong.

Every time you make a phone call, software in the phone system checks to see if the number you’re calling has been ported. Since phone numbers are geographically assigned, there is a shared porting database for each calling area in which the calling switch looks up the dialed number (DN) to get the routing number (RN). If the number hasn’t been ported the DN is the same as the RN, but if it has, RN is a number assigned to the switch to which the number has been ported. Then the call is routed based on the RN, but it also sends along the DN so the target switch knows who the call is for. The shared databases are run by a neutral party (Neustar in the US) and every telco pays to support it. The system was designed this way so that numbers that have been ported away don’t put an extra load on the “donor” system from which it was ported.

Email doesn’t work like that. There is a DNS lookup for the domain name, the part of the address after the @ sign. but all mail within the same domain is routed to the same place. For the small minority of Internet users who have their own domains, they can change the domain’s DNS records to change where the mail goes, but for users who get their addresses from their ISP or their employer, it’s tied to the ISP or the employer. You can imagine a system in which every mail delivery did a DNS lookup of the e-mail address first, but that’s not how the mail system works.

But since this is a government mandate, is there any way to make this sort of work?

There were two other approaches for phone number portability proposed and discarded, call release and call forwarding. In call release, the call first goes to the original switch, which sends back a status message saying the number has been ported to another switch, and the calling switch then reconnects to the other switch. Call forwarding should be familiar to everyone—the called switch places a call to the real destination switch and connects the incoming call to it.

E-mail has analogs to both of these. For something like call release, the SMTP standard has always had a status code that a recipient system can send back to a sending system to say that the recipient has moved, and giving a new address. As far as anyone can tell, nobody has ever used that code, but it’s there if anyone wants to give it a try. Mail forwarding, on the other hand, is very common.

The least awful way I can think of to make something like this work for email is that the user’s new provider can contact the old provider on the user’s behalf, and request the address be forwarded. So long as it’s forwarded, the new provider pays the old one a modest monthly fee, mostly to give the providers an incentive to cancel the forward when the user leaves. The fees would probably net out in most cases so the costs would be mostly administrative.

Mechanically, that kind of setup would not be very hard. Administratively, it would be a nightmare. If the forwarded mail starts to bounce are they allowed to turn it off? Does the old provider do its usual spam filtering? (What if the user left because the filtering was lousy and lost a lot of real mail?)

Another possibility would be for the old provider to keep mail accounts active even though the account is otherwise turned off, and let people pick up mail from its mail server. This is surprisingly common now, often by accident. For example, I cancelled my BT broadband account in July when I left England, but the associated mail account still works, seven months later. Mechanically this still isn’t hard, but if it’s a required service, now each ISP now has a permanent obligation to provide mail service to people from whom they no longer get any income, and with whom they have no other relationship. How do they know when to turn off the mail? If the user doesn’t pick up the mail for a month? Six months? A year?

So my main advice is to forget it, since there’s little evidence that this is a service so important it needs to be mandated. On the other hand, ISPs might find a small new income stream by selling forwarding service, like many post offices do. If the user is willing to pay $20/yr, that’d probably cover the cost of keeping a mailbox open, and would solve the problem without having to invent new rules and mechanisms.