3 ms·

> The root problem of spam is not that email is cheap to send. I spent some time searching the web and even dug through your comment history but am failing to

by saurik 3y ago

> The root problem of spam is not that email is cheap to send. I spent some time searching the web and even dug through your comment history but am failing to find where this might be; would you be willing to provide a link to help my deficient searching ability? I will note that I did find you specifically claiming that spam is all about "speed/volume", and it seems like most of your discussion of what to do is to increase the cost of sending lots of email. Honestly, you seem to be arguing a straw man here, which is really frustrating as, I get it: you are an expert in the field of spam... yet, your conclusion seems like it is still incorrectly analyzed :(. > There are enormous numbers of bulk emailers that are useful to their subscribers. ... Only rich western elites could look at email and conclude that the best improvement is to charge lots of money for it. No one credible is trying to do this: every complete and in-depth model for this that I have heard involves variable pricing, decided by the receiver per-sender, and entirely solves these use cases. > Jameson is strongly implying here that email would somehow be better if it was frozen in time in 1991, with the same number of users - a platonic ideal of decentralisation in which the only participants are, uhm, the military, universities and corporations, who interconnect only with other similar institutions and only grant email accounts to employees. I honestly just can't find evidence that this is the case: even the old hashcash project talks about these issues and posits solutions for them... maybe those solutions don't work, but you have to argue that.

mike_hearn 3y ago
Hi, I wrote about it before here: https://moderncrypto.org/mail-archive/messaging/2014/000780.html https://moderncrypto.org/mail-archive/messaging/2014/000780.... > most of your discussion of what to do is to increase the cost of sending lots of email. I don't think I've ever proposed increasing the cost of sending lots of mail? Or not that I recall! I've definitely discussed increasing the cost of sending lots of unwanted mail, but that's not the same thing. > No one credible is trying to do this: every complete and in-depth model for this that I have heard involves variable pricing, decided by the receiver per-sender, and entirely solves these use cases. Variable pricing doesn't solve the issue because a key part of email is the ability to receive emails from strangers who haven't contacted you before. This can also be the case for bulk mailers, e.g. if you write down your email address with pen and paper for your local school, who then emails every parent. How do you price that? And is this "price" measured in abstract units people have no intuition for, like CPU cycles, because it's surely not in dollars. So I've never seen a workable proposal for introducing pricing or charging into email. > maybe those solutions don't work, but you have to argue that. Jameson only mentions hashcash in passing, he's mostly writing about how email was better when it was more decentralised, before the coming of big providers. Which is a bit odd to me because in the early days of the internet the only people with email accounts were those hosted by big providers. Maybe he doesn't view them as such because they weren't primarily in the email business. I've argued why hashcash doesn't work above, and I don't know anyone in the email industry who has ever taken this idea seriously except Bill Gates, and that was decades ago. The email industry knows how to keep spam under control, and it's with decentralised systems like reputation. Now Jameson doesn't consider reputation to be decentralised, but this is highly arguable. I view it as very decentralised. Every email server can compute reputations to whatever degree of detail they desire. Unlike pre-reputation filters there's no need to haggle with ornery blacklist owners or sendmail admins over what exactly the word "spam" means. Reputations are entirely under the control of both sender and receiver, with nobody in the middle, and so you can't get more decentralised than that. I suspect what he objects to is that whilst the reputation system is more decentralised in the abstract, in practice there are so many potential email senders that a small site can't actually compute them. The problem here isn't machine resources by the way, those are cheap. It's that because receiving email from strangers is such a common and desired use case, most email comes from a sender your server never saw before, both good and bad. The old DNS BL system was more centralised but the BLs were openly available, so it may have felt decentralised just because you could run your own server and that's what tech people tend to think of when they hear the word. But that server wasn't really yours in the truest sense, because you'd allowed some random third party to decide what mail you could see (and incidentally exposed the domains that were emailing you to the BL operator, a big privacy leak). Ultimately, the current email network works fine. Its problem is neither centralisation (there are lots of competing email providers) nor the way spam filtering works, which the big players have systematically improved for everyone. Its biggest problem is simply that as a highly decentralised system it's basically impossible to improve, and so gets regularly outcompeted by proprietary platforms. One day it'll go the way of IRC and we'll all move to a centralised platform. It's pretty much already happened for personal communication, which moved to WhatsApp and Facebook.

by saurik 3y ago

mike_hearn 3y ago

Hi, I wrote about it before here: https://moderncrypto.org/mail-archive/messaging/2014/000780.html https://moderncrypto.org/mail-archive/messaging/2014/000780.... > most of your discussion of what to do is to increase the cost of sending lots of email. I don't think I've ever proposed increasing the cost of sending lots of mail? Or not that I recall! I've definitely discussed increasing the cost of sending lots of unwanted mail, but that's not the same thing. > No one credible is trying to do this: every complete and in-depth model for this that I have heard involves variable pricing, decided by the receiver per-sender, and entirely solves these use cases. Variable pricing doesn't solve the issue because a key part of email is the ability to receive emails from strangers who haven't contacted you before. This can also be the case for bulk mailers, e.g. if you write down your email address with pen and paper for your local school, who then emails every parent. How do you price that? And is this "price" measured in abstract units people have no intuition for, like CPU cycles, because it's surely not in dollars. So I've never seen a workable proposal for introducing pricing or charging into email. > maybe those solutions don't work, but you have to argue that. Jameson only mentions hashcash in passing, he's mostly writing about how email was better when it was more decentralised, before the coming of big providers. Which is a bit odd to me because in the early days of the internet the only people with email accounts were those hosted by big providers. Maybe he doesn't view them as such because they weren't primarily in the email business. I've argued why hashcash doesn't work above, and I don't know anyone in the email industry who has ever taken this idea seriously except Bill Gates, and that was decades ago. The email industry knows how to keep spam under control, and it's with decentralised systems like reputation. Now Jameson doesn't consider reputation to be decentralised, but this is highly arguable. I view it as very decentralised. Every email server can compute reputations to whatever degree of detail they desire. Unlike pre-reputation filters there's no need to haggle with ornery blacklist owners or sendmail admins over what exactly the word "spam" means. Reputations are entirely under the control of both sender and receiver, with nobody in the middle, and so you can't get more decentralised than that. I suspect what he objects to is that whilst the reputation system is more decentralised in the abstract, in practice there are so many potential email senders that a small site can't actually compute them. The problem here isn't machine resources by the way, those are cheap. It's that because receiving email from strangers is such a common and desired use case, most email comes from a sender your server never saw before, both good and bad. The old DNS BL system was more centralised but the BLs were openly available, so it may have felt decentralised just because you could run your own server and that's what tech people tend to think of when they hear the word. But that server wasn't really yours in the truest sense, because you'd allowed some random third party to decide what mail you could see (and incidentally exposed the domains that were emailing you to the BL operator, a big privacy leak). Ultimately, the current email network works fine. Its problem is neither centralisation (there are lots of competing email providers) nor the way spam filtering works, which the big players have systematically improved for everyone. Its biggest problem is simply that as a highly decentralised system it's basically impossible to improve, and so gets regularly outcompeted by proprietary platforms. One day it'll go the way of IRC and we'll all move to a centralised platform. It's pretty much already happened for personal communication, which moved to WhatsApp and Facebook.