3 ms·

I hope he has a good spam filter... I know pg pioneered bayesian spam-filtering, but I wonder how well it copes with publicly known email address? I tend to as

by 13ren 18y ago

I hope he has a good spam filter... I know pg pioneered bayesian spam-filtering, but I wonder how well it copes with publicly known email address? I tend to assume Gmail has the best spam filtering (with their massive database of email), but I still get about 7 per day that it's not sure of, and my account name isn't very publicized.

rms 18y ago
Do you click the report spam button?
- 13ren 18y ago
  The 7 per day that it's not sure of are already in the spam folder. On the rare occasion one gets to my inbox, yes I click the report spam button.
  - rms 18y ago
    So it works... your spam goes to the spam folder.
    - 13ren 18y ago
      Yes, it does. I'm a bit confused though - I wasn't in question about that. Let me clarify: I was wondering if pg's bayesian filter works so well that publishing his email address won't cause a bump in spam for him - since he seems to hide it (as I do mine). I was using Gmail as one example of bayesian filtering. Mainly, I was gently suggesting it's better to not publish anyone's address here where it will get harvested. BTW: When I think about it though, Gmail didn't work as well for me in the past.. but I don't know the cause of the improvement: it could be because I've been a lot more careful since then; or because Gmail has improved in general; or it's learnt for me specifically. I think a big factor is my old uni accounts stopped being diverted to it, and they used to catch a heap of spam. That's probably it.
davidw 18y ago
I get a relatively high number of false positives, but I admit that my situation is probably an outlier. I get error reports, mailing list moderation messages, and other stuff that must be hard to filter on.
jodrellblank 18y ago
It would cope brilliantly. Because PG gets that spam is a personal definition (http://www.paulgraham.com/better.html http://www.paulgraham.com/better.html - last two par's before notes). Running a spam filter at the level of a company, or several companies, will never work properly. Ours doesn't, despite running several popular anti-spam engines together. It's good compared to not running any filter, but it can't be tuned with enough granularity, and it doesn't learn. It's a combination of "crummy but available is good enough", failure to believe/see that it could be better, and Yudkowsky's "not holding ourselves to a high enough standard". GMail could do it, since they have access to not only all your mail, but also the possibility to identify "mail arriving at x many accounts system wide".

by 13ren 18y ago

rms 18y ago

Do you click the report spam button?

13ren 18y ago
The 7 per day that it's not sure of are already in the spam folder. On the rare occasion one gets to my inbox, yes I click the report spam button.
- rms 18y ago
  So it works... your spam goes to the spam folder.
  - 13ren 18y ago
    Yes, it does. I'm a bit confused though - I wasn't in question about that. Let me clarify: I was wondering if pg's bayesian filter works so well that publishing his email address won't cause a bump in spam for him - since he seems to hide it (as I do mine). I was using Gmail as one example of bayesian filtering. Mainly, I was gently suggesting it's better to not publish anyone's address here where it will get harvested. BTW: When I think about it though, Gmail didn't work as well for me in the past.. but I don't know the cause of the improvement: it could be because I've been a lot more careful since then; or because Gmail has improved in general; or it's learnt for me specifically. I think a big factor is my old uni accounts stopped being diverted to it, and they used to catch a heap of spam. That's probably it.

davidw 18y ago

I get a relatively high number of false positives, but I admit that my situation is probably an outlier. I get error reports, mailing list moderation messages, and other stuff that must be hard to filter on.

jodrellblank 18y ago

It would cope brilliantly. Because PG gets that spam is a personal definition (http://www.paulgraham.com/better.html http://www.paulgraham.com/better.html - last two par's before notes). Running a spam filter at the level of a company, or several companies, will never work properly. Ours doesn't, despite running several popular anti-spam engines together. It's good compared to not running any filter, but it can't be tuned with enough granularity, and it doesn't learn. It's a combination of "crummy but available is good enough", failure to believe/see that it could be better, and Yudkowsky's "not holding ourselves to a high enough standard". GMail could do it, since they have access to not only all your mail, but also the possibility to identify "mail arriving at x many accounts system wide".