View Single Post
  #34 (permalink)  
Old June 24th, 2003, 09:21 AM
David Long David Long is offline
Authorized User
Join Date: Jun 2003
Location: , , .
Posts: 11
Thanks: 0
Thanked 0 Times in 0 Posts

Please accept my apologies for taking this thread off-topic, but I, Dan and Stephen thought it best to try and clear up things that had already been mis-reported.

Back on topic: Many suggestions have already been given in this thread, but some of the problems seem unsolvable:
  • Authenticating users by something other than the "from" address - not difficult when dealing with replies (the authentication code would just be included in the reply) but new threads started by email would need something extra included in the body, which people just won't remember. We tried reminding people on the old site to cut down on the amount they quoted and to not quote headers/footers, but it just doesn't work, people just want to type their message, send it and forget about it - they don't want to have to take extra steps just to post to a mailing list, especially if they don't have to do that on other mailing lists they use.

    The "registered IP address" mechanism probably won't work reliably in most cases due to SMTP relays (this will affect most users to some extent, I'd have thought), and insisting on PGP signatures will alienate users who currently don't use them. I can't see an alternative, workable solution for this, and nobody else has posted one in this thread either.
  • Detection of overquoting is never going to be foolproof. I tried many times to write some code that could handle any quoting mechanism so the archive would be a bit neater, but I gave up in the end as I figured it's just not possible - there's simply too many mailers out there that do it in different and incompatible ways.

    If you really want to achieve this, I can't see how you'll do it without a human operator in the loop to figure out exactly which bits of the original message are relevant, and which bits aren't. A computer program would basically need full understanding of the English language in order to be able to do this successfully, even if you can figure out which bits of a message are quotes and which bits are new text - say you know you've got 1000 lines of quoting and 10 lines of new reply in a particular message, which of the 1000 lines do you keep, if you can't understand the context?
  • Figuring out replies to topics and attaching them to the correct threads is technically possibly by subject line (this is how the old web interface list browser and archive worked - strip "Re:" and similar prefixes and all whitespace to give you a unique topic string), though it's not 100% foolproof. It did work well enough for us, most people use unique enough subject lines to be able to figure it out.

    Alternatively, you could include a unique identifier at the start of the subject line, which would be intact in the reply. You could even use this to implement true threading rather than all replies being attached to a single root-level topic, but as far as I can see Snitz doesn't support this anyway.
  • Out of office and other automated replies can mostly be blocked by scanning incoming messages for known substrings, but again this won't be 100% perfect. There'll always be someone with an out-of-office reply in a foreign language that doesn't get trapped with this method.
  • Spam filtering can be achieved by placing SpamAssassin, a Bayesian filter or some similar software between the receiving mail host and the software that puts the message up on the forum. Again, this isn't going to be 100% perfect; determined spammers will always find a way around automated checks eventually (if they couldn't, spam wouldn't be a problem for anyone!). But, like we've already said, spam wasn't a problem on the old forums, and with a few extra checks in place, I can't see it being a problem on the new ones either.

As has already been said, the easiest and cheapest way to integrate this would be to use a mailing list manager that someone else has already written, then try and write an interface to bolt it into the Snitz forums, but if Wiley don't want to use a third-party product, then I'm stumped as to how to proceed further with this, especially the authentication requirement.

Finally, credit must be given to Hal for talking to Wiley and trying to make the forum usable for the masses by taking it back to the ways that everyone was familiar with. I don't think I'd have taken on that job :)