inspired by Rands

[dsrtao]

Let's take the spam problem on by being aggressive about it.
People are dumb. If they see spam, some small percentage is going to buy from it. Therefore they shouldn't see it.

Every mail program knows how to filter. None of them (except, kinda sorta MH) is built around the idea that filtering and sorting is the most important thing. That's what we need.

I call the interface Buckets. It could be built on top of Thunderbird or GMail pretty easily. Here's how it works: your home screen is a set of icons that include counters of new mail. The icons are Friends, Family, Work, Hobbies and Slops. The user cannot change these. At the beginning of a user's experience with Buckets, all mail comes into Slops. The Slops bucket must be sorted; first by hand, later automatically. You cannot reply to a message in Slops: you must assign it to one of the other buckets or consign it to the Burn Bin. The buckets remember salient features of what you put into them - Subject, From, List headers and so forth. It might be a good use for CRM-114. Whenever confidence is high that a new message fits into a non-Slops bucket, it is automatically assigned. Spam is recognized as automatically as possible, here, and flagged.

Advanced users can create subsidiary Buckets inside the big ones. Mail is assessed for suitability in a subBucket only after it has qualified to escape Slops.

Every message has a Change My Bucket control.

Buckets stop showing older messages, but you can search for them or star them to stay on top. Messages are never deleted except by an explicit Kick The Bucket.

Search can happen over all Buckets or just the one you're in.

Now the radical part: automatic whitelisting and blacklisting. Moving a message into a Friends/Family/Work/Hobbies Bucket automatically whitelists the sender; moving a message back into Slops removes the whitelist, and moving it into the Burn Bin is a signal that it's spam or malware, and the source is considered suspect.

Comments

[mangosteen]

It's intriguing and definitely a good first step, but I don't see how it solves the problem. In the end, the user has to look in Slops to see what should get classified. If they really want to buy v1agra or c1@lis, then they'll just file the message into Hobbies and then respond.

It feels like it immediately goes back to the problem of "a small amount of people will always respond."

[dsrtao]

There's nothing you can do about willful stupidity except get into issues of privacy violation that I'm not interested in aiding (i.e. reporting keeping elephantiasis spam as a sign of deviancy).

Also, see edits.

Also, anyone who claims to stop the spam problem outright is deluding themselves.

[goldsquare.livejournal.com]

Can you factor "the wisdom of crowds" into it, somehow?

[dsrtao]

http://razor.sourceforge.net/ does that. SpamAssassin, among others, incorporates it.

[goldsquare.livejournal.com]

Oh, I know.

But where would that model appear in your buckets model?

[n5red.livejournal.com]

CRM-138? I haven't heard of it. Any relation to CRM-114?

[dsrtao]

It's hard to do without leaking privacy. Sending off a hash of a suspected spam doesn't leak much, but if you send off hashes of everything you receive, hoping that someone has already categorized it into a useful bucket for you (and only the top level buckets are the same for every user) is wasteful. A shiny nickel says that CRM-114 http://crm114.sourceforge.net/ can do it faster, locally, given a reasonable working corpus.

[dsrtao]

Typo, changed just before I saw your message.

[goldsquare.livejournal.com]

Well, my initial gut reaction is that users won't have much a privacy affection for SPAM, and wouldn't mind aggregating or sharing it.

But maybe that isn't a good idea. For example, I've tried to unsubscribe to a couple of commercial mailing lists (notably Disney) with poor results. I've marked them as UCE. And they are: I asked for it, I asked not to get it, I am getting it 6 months later.

But if you are actively planning a Disney vacation, and my sharing my information means your data hits the bit bucket, you won't be pleased.

Would your model permit the granularity of marking something as "everyone should think this is garbage" and "I think this is garbage", along with some tolerance of bad actors with a grudge against Disney?

[dsrtao]

I wouldn't suggest cooperative selection of buckets; for spam there is already Razor, and for anything else I don't want to leak that (for example) everyone ranked director and higher in such and such corporation just got the same email.

[robertdfeinman.livejournal.com]

Thunderbird has a feature where messages with links in them that don't match the visible URL are flagged as a possible scam.

I think guessing at malware categories while leaving the message in the incoming bucket might help people be more thoughtful.

[dsrtao]

The Slops Bucket isn't "incoming"; all the Buckets are incoming, for whatever messages get matched automatically.

Slops is where unclassifiable messages go -- it's where the system isn't sure, and needs human intervention. If you can put a message into a malware category, it sits in the Burn bin.

[jducoeur]

The hard question: why would "I" (the naive ordinary user who is the problem here) use this? More particularly, why would I use this instead of what I already use? This appears to be more work, and therefore less convenient, for me.

If you start with "all mail comes into Slops", you immediately lose on that score -- most users will look at it, say "This is too much work", and go back to something else. Arguing that it's going to be less work in the long run (much less that it's for the good of society) isn't going to get you very far. If it's going to succeed, it has to be clearly at least as good (preferably better) on the *individual* level...