Dan Weber <[EMAIL PROTECTED]> said: [with respect to subaddress handling] > I reviewed this, however, people may use different symbols to show, like > '-' is accepted and '+' as well. The former is more common. This could > create a mess with usernames that have a '-' in them. On the discussion > on how to support this, it should be handled the following way: Link > aliases to appropriate mailboxes.
Linking aliases to specific mailboxes is a very good idea, I think we should put that on a 2.1 or 2.2 todo list, along with the proposal to separate forwards from deliveries (forwards as varchar, deliveries as bigint). An interesting combination of the two might allow semantics of passing a message through a filter and then delivering it to a user. Limited, but useful. Anyhow, let's table this until 2.1 or 2.2 and put a note in the TODO file. I have not seen any subaddresses using '-', unless you count mailing list managers... but those typicall do their own address parsing, so I don't think that DBMail needs to worry about it. Have you seen it used elsewhere? The Sieve Subaddress Extension (I forgot the RFC, but it's in earlier posts) suggests using # as one possibility, for example in an email-to-pager gateway. Of course, it then goes on to specify the BNF grammar with '+'. Go figure. I'm beginning to think that we should just roll 2.0rc8 as it is -- there's been a lot of work already that needs to be tested closer to production. I've spent a whole lot of time thinking about how to solve the mailbox issue, and except for an MTA filter, there's no easy solution. It might take me a week to finish up the code I started writing a few months ago that rearranges the pipe/sort interface and allows the sort subsystem to receive more detail about the delivery. This code and more will be required for 2.1 when sorting is actually implemented, so it won't hurt to get a little head start on it to help fix subaddress security. Aaron
