Fog Creek Software
Discussion Board

Email Postage

I've been thinking about email postage system for a while, and it's hard for a number of reasons. I can think of solutions but stopped when I heard of a much better ideas.


Post a bond ($100? $1000?) and your account can send email. Inifinte amounts of email--what you've just paid for is to be added to a global white list. Fail to obey the rules and you get removed from the list and forfeit your money.

This still lets automatic whitelist software (bounce replies and ask for human response to be added to the 'OK' list) work, but also allows mail lists (no time to read the bounced replies) to work just fine.

Let the cost of the white list be $100/year/ISP and people will maybe even go for it.

Here's a link from Google, haven't read it.

Thursday, November 14, 2002

This begs the question: who decides if you "failed to obey the rules?"

If the bond is $100, it's no problem for a spammer. If it's $1000, nobody is going to go for it.

The benefit of one penny mail is that as long as you pay a penny per message, you're obeying the rules by definition.

Joel Spolsky
Thursday, November 14, 2002

Sounds OK to me, I must be a small email user, my 'sent' folder contains 469 items and I've been using this PC for 2 years now. $4.69 - I can afford that.

Something must be done to destroy spam.

Thursday, November 14, 2002

Interesting.  I just checked my mail folders.  I've only ever received two messages from Joel On Software -- on Dec. 12 and Dec. 25, 2001.  I'm pretty sure that my e-mail provider doesn't do any filtering.  Are the messages being blocked somewhere earlier?  Or, maybe my e-mail address simply got dropped from the database somehow.

Alex Chernavsky
Friday, November 15, 2002

I knew I've heard this before:,10738,2799926,00.html

Friday, November 15, 2002

email postage might not be necessary, paul graham and a few others claim that simple bayesian filtering gets rid of 99.5% + of spam with next to no false positives.

I've downloaded this thing, and am trying to integrate it with my web server. it will be interesting to see if the theories are correct. (there are a number of people who say it "works for them")

Will C
Friday, November 15, 2002

The hardest problem, by far, with a postage system is adoption.
I can design a fully distributed system which allows multiple payment systems, and you can transmit (via email or other protocol) a negotiation every time you send a message, the transfer of payment to the recipient with a cut to the processing agency, and so on.

The issues are adoption and the cost: can I charge $1 or $10 or $100 to send me email because I've been the target of hate mailers (not just spam) for years and years? If it's only in true micropayment land ($.01), the management cost is higher than the payment.

But really, bonding is only for people who manage mail lists. It's not a big deal to post $1000 for lots and lots of lists. And if you have time and no money, you can deal with the automated whitelist systems described above.

Also I believe bonds aren't necessarily paid up in full. Can a financial person describe them more? My understanding is that you pay out, say, 10%, but have a debt of the other 90%. If you leave the system, you get your 10% back and the 90% is forgiven. The percentages can vary based on your credit rating.

The system I describe above seems to charge your bond based off of complaint volume. Once you can track the spam (which the bonding ensures), a lot goes away immediately because you can do something to the sender.

Of course harrasment via complaint filing is possible, but that's a risk to take.

Friday, November 15, 2002

And remember, inexpensive amount in US or other developed nation is not so small amount in poorer nations.

anti WTO
Friday, November 15, 2002

Another idea I've seen proposed is payment in CPU or human time:

Friday, November 15, 2002

I really don't mind paying 0.01CAD per mail. I think the rate should be even higher. Say, 10% of snail mail's, that is about just 0.05. Let's start with Hotmail? :)

Friday, November 15, 2002

one line of enquiry is "why does spamcop blacklist".  Maybe Joel's trust in them is misplaced?

Whilst I like the idea of stamps for email, I just don't see it catching on because the only person I trust to actually do something about the spam I receive is /me/.  And that still chews up my bandwidth.

Bayesian filters seem to do a very good job; on the other hand, sending a footnote about fogbugz being released might trigger some bayesian alerts ;-)

Friday, November 15, 2002

You might want to check out Amy Wohl's travails in her weblog entry 'SPAM and IRRESPONSIBLE ISPs'

Friday, November 15, 2002

Why not just stop sending email?  Hear me out.  I've been watching Livejournal -- -- lately, and I have noticed something interesting: People use it as a means of communicating.  Someone posted directions to their house and "friends' locked" message for privacy.  It strikes me that this, or an version of this with Groove (?), might easily replace email.  Instead of an open message spool waiting for spam, people could setup private groups where "handshakes" were required.  Setup your "handshake key" on your website, not unlike PGP's code blocks, and if someone wanted to email you they would send a request.  Sure you may get deluged with spam requests, but you could filter those into a "request" folder and delete them en masse or scan through quickly for people who you recognized -- I assume people could/would post "handshake keys" on messageboards, newgroups, etc. or pass them in IM's and chats.  Just a thought.

Basically, though, I just dislike the idea of paying for email because it seems to be a quick way to eliminate a lot of discourse.  Imagine if you charged a penny for every visit to your website to keep RSS pullers from tripling your bandwidth, you'd quickly lose a lot of potential customers who didn't want to pay to look at something they might want to buy.  I may be insane, and I am certainly not as knowledgable as most here, but paying for email just sounds off base.

Andrew Burton
Friday, November 15, 2002

Regarding the one-cent-er-e-mail idea: yes, yes, yes, yes, yes.

Hardware Guy
Friday, November 15, 2002

An interesting idea is to use a cryptographic hash.  Delivering an email to a user could require a bit string whose hash has certain properties.  For instance, I could declare that in order to send me an email, you have to produce a string of bits whose SHA hash ends in 0xDEADBEEF (or whatever).  If you think you are getting too much spam, just increase your requirements.  There would be no burden on a friend sending you an email - once he types your address into the "to" field, his machine could start looking for an appropriate bit string.  Even if it takes all of a minute, it wouldn't be a burden, but should be enough to stop all but the most targetted spam.

Friday, November 15, 2002

My ISP uses something called SPAM-Guard.
It holds suspected spam on the server and then lets you know how much
spam is waiting for you every few days, so you can log on, check it and delete. I get about twenty spams a day, (which is not much I know) and less than 1% of it gets through. It's been months since it stopped any real mail getting through, although I don't get many emails from people I don't know.
If you got lots of unsolicited genuine email it might be too agressive for you. for more info

(And I don't get paid for saying this. Just a satisfied customer)

Spamless in Ontario
Friday, November 15, 2002

I think spam proliferation is a manifestation of a divide-by-zero error.

Spammers sent as many junk emails as their budget allows - total emails sent= budget/cost per email. As the cost approaches zero, we trend towards infinite spam. Joel's proposal essentially ensures the denominator is non-zero!

I like it - it addresses the cause of the problem. Spam filters address the symptoms.

Matthew Thornhill
Friday, November 15, 2002

I thought of charging for email and had wondered if there was something patentable about it. Oh well, I guess not if hundreds of other people had the same idea. The difference with my idea is that you would be able pay as much as you want to send the message, and the receiver would get paid a related amount to read the message. For example, if you get 10 cent spam sent to you, you get 5 cents to read it.  If you want to get a message to a busy Hollywood Producer, send him a $100 email. He gets $99 when he reads it.

Important to note that you don't have to "boil the ocean" with this idea. This doesn't have to cover all email. Just a few hundred users would get utility of it. It could sort the mail into folders by paid amount. After awhile you could ignore the unpaid spam and read the paid spam to get a few cents like using a Discover card.

Fred Brunet
Monday, November 18, 2002

*  Recent Topics

*  Fog Creek Home