The Portable Freeware Collection Forums

Hydaral wrote:

m^(2) wrote:Just fount this:
http://highscalability.com/blog/2011/2/ ... 00-re.html
It may be worth trying.

That sounds good in theory, but spam detection is not that hard if you don't have to worry about false positives, from that webpage:

Mollom does not allow you to present an example of a wrongly rejected chunk of content that should be accepted

It's obvious to do this that way. Otherwise bots would learn to always appeal and the system would be useless.
And I'm pretty sure they do worry about false positives and worry a lot, otherwise they would end up dead very soon. It is hard to keep both false positives and false negatives low. And even humans aren't perfect at it. About a moth ago I reversed my decision to ban one person because I thought it was a spam bot.

m^(2) wrote:

Hydaral wrote:

m^(2) wrote:Just fount this:
http://highscalability.com/blog/2011/2/ ... 00-re.html
It may be worth trying.

That sounds good in theory, but spam detection is not that hard if you don't have to worry about false positives, from that webpage:

Mollom does not allow you to present an example of a wrongly rejected chunk of content that should be accepted

It's obvious to do this that way. Otherwise bots would learn to always appeal and the system would be useless.
And I'm pretty sure they do worry about false positives and worry a lot, otherwise they would end up dead very soon. It is hard to keep both false positives and false negatives low. And even humans aren't perfect at it. About a moth ago I reversed my decision to ban one person because I thought it was a spam bot.

They quote spam classification percentages, but not false positives though. I could make a system that stops > 99.9999% of spam just by letting through 1 in every million emails. Of course my false positives would be terrible.

Hydaral wrote:

m^(2) wrote:

Hydaral wrote:That sounds good in theory, but spam detection is not that hard if you don't have to worry about false positives, from that webpage:

It's obvious to do this that way. Otherwise bots would learn to always appeal and the system would be useless.
And I'm pretty sure they do worry about false positives and worry a lot, otherwise they would end up dead very soon. It is hard to keep both false positives and false negatives low. And even humans aren't perfect at it. About a moth ago I reversed my decision to ban one person because I thought it was a spam bot.

They quote spam classification percentages, but not false positives though. I could make a system that stops > 99.9999% of spam just by letting through 1 in every million emails. Of course my false positives would be terrible.

Yeah, sure, the false positive rate alone is useless. But f.e. the list of their clients is not.

OT: I just got "You may embed only 3 quotes within each other."
I think that the forum should either accept more than 3 (I think such limitations are wrong) or trim the quote depth by itself.

I think limiting quotes within quotes makes sense otherwise you end up messing up the forums look or with really small quoted areas.

carbonize wrote:I think limiting quotes within quotes makes sense otherwise you end up messing up the forums look or with really small quoted areas.

And by limiting them you mess up the quotation history. The problem with the first is that you have to scroll more. The problem with the second is that sb. reading it later may get lost, especially when reading quickly and when there's a lot happening in the thread.

I think limiting the nesting to three levels is more than adequate in most situations to give you a feel for the back story.

On another forum I noticed that the storm seems over, from several bot accounts per day it went down to 1 in last 10 days (possibly some unidentified though). Is it the case here too?

As far as I am aware nothing has shown on the forum for the last 10days too.

m^(2) wrote:On another forum I noticed that the storm seems over, from several bot accounts per day it went down to 1 in last 10 days (possibly some unidentified though). Is it the case here too?

guinness wrote:As far as I am aware nothing has shown on the forum for the last 10days too.

Glad you both noticed. Andrew hasn't made a formal announcement about how we're handling that so I'll let him decide if he wants to jump in and explain the system.

There has been more entry spam, but lots of eyes on the site have frequently killed those without moderator involvement.

m^(2) wrote:On another forum I noticed that the storm seems over, from several bot accounts per day it went down to 1 in last 10 days (possibly some unidentified though). Is it the case here too?

I don't know if it's related, but I was getting about 10-20x as much spam in my personal email as I used to, now it's dropped back to 1 or 2 a day. Maybe a new spam house started up and then was shut down just as quickly.

Well phpBB 3.0.8 now has the ability to look at ip blacklists and block registration attempts, this should for a while help a bit.

Zach Thibeau wrote:Well phpBB 3.0.8 now has the ability to look at ip blacklists and block registration attempts, this should for a while help a bit.

Sadly, many bots are on computers with dynamic IPs. It's a troublesome soultion...but still worthwhile.
BTW in a meanwhile spam level on the forum that I'm moderating bounced back.

Well I know for sure the moderators and Andrew are doing their best, just wish I was back on the team sometimes.