not claiming a bug, just don't remember ever setting my posting preference to "CODE". though if that's a legitimate setting, why does it bug people that it's in use and why don't more people use it?
i changed it to plain text, because I don't like controversy.
nope, we're counting all reasons for rejections in those figures reported, we don't block by IP except on temporary basis or what's in well established RBLs, and listings in those RBLs are all temporary (no use of permanent RBLs like that one particular one who blackmails people into paying $50 to get their IP off the list after baiting senders with subscriptions)
Let's say you notice 10.10.10.0/24 has only ever sent junk, why not block the entire class C?
If a reliable reputation/feedback database says that mail originating in that subnet is reported as junk 100% of the time, why allow that subnet to waste your money?
sorry, just hit reply, and that's the font that came up after preview/submit. I'm not normally a LOOK AT ME!!! type of guy. Well, I am. Just in this case it was inadvertent.
long term, we've been allowing into the environment roughly the same volume of email per customer for 10 years. Some spam gets through, most does not, and there are few false positives. those that are labeled false positives are most often bulk mail that people mark as junk. So IMO, it's junk mail.
We use rules at the protocol level, DNS responses, RBLs (combined into one large RBL with miltiple return values), external reputation lists, internal dynamic reputation lists, rate limitations, and multiple feedback systems to provide this level of protection, that's before content filtering and personal white/black lists.
Just today, on the protocol layer, we're blocking 60% at banner (RBLs, bad DNS) , %14 of the remainder at HELO, %3.5 of the remainder Mail From (fake domain names) and finally a good chunk of what's left is blocked because it's destined to bad email addresses (which feeds back into the reputation lists).
Customer feedback helps stop those who are newly spewing spam, and since the feedback systems are widely distributed over many different email service providers, a massive spike at one translates into a blocked email at the others (whether by IP or content).
Better still, we do the same thing on the outbound side of things. If a customer catches a virus, they're cut off from email pretty fast and the feedback system is a very very tight loop internally.
But you are right, it's an ever escalating war, and if we could skip a few steps and jail (permanently, with broken hands) the spammers and bot coders, we wouldn't have to spend the money on the filtering and RBLs and feedback loops and hardware. We adjust the rules slowly over time, the feedback systems are maintained by the "trusted" customer, we're spending hundreds of thousands of dollars a year to protect against junk mail. I'm not certain of the math here, but an educated guess, this translates to around %5 of the cost to serve a user's mailbox. That's just operations staff time, and datacenter space for the extra hardware, the hardware itself, the subscription fees to the antispam service, wasted bandwidth etc.
you are correct, the missing data point is the volume of email considered "not spam". This line in the graph stayed the same over the range, or within a minor fraction of a percent of the same. it's the spam counts that have dropped since 10/2010. The customer base also represents a large number of domain names, hundreds of thousands of domain names. One of our largest customers has been offering email since 1995, with many accounts in their domain being around for over a decade. I think it's a pretty solid sample of email accounts.
my graphs show a steady decline in spam capture rates since October, 2010. we're measuring an average daily rate about 1/2 of this time last year. (millions of mail boxes, dozens of MX servers, decent antispam filtering) We're blocking around %91.2 of mail at the perimeter as opposed to %98.8 last year.
I work for an email service provider, we're catching many each day, most less than 500 emails at a time. I think about 1/2 of them are compromised PCs as they're using the same IP addresses the customers use, different HELO hostname and all that but they're still authenticating from the same place. That's the wild part. I watched a network sniff play out on screen, showed the authentication stuff, same user ID and password, different HELO hostname and headers, right along side another session where the user was sending legit email.
The other portion are clearly phished accounts, customer in Boise, connection from China for example.
The kicker is that we've had to turn off our internal reputation system based on the age of the email account. Used to be > 1 month old had higher limits than < 1 month old (for the love-em-then-leave-em accounts), but today, no one is trusted.
The only good thing is they seem to come in phases, where a particular campaign of the exact same email comes from dozens of accounts, for hours at a time, then switches to a new campaign later. Makes filters easier to manage.
If you're not paying for it, they can do whatever they want with it, how else are they supposed to make money off the customer for providing the service?
Don't like it, change your email address.
They'll wind up like Excite, for example. Too many freeloaders, not enough clicks on ads.
they get unexplained skin growths from crapping in their suits because they're too hard core take the time to find a port-a-potty and risk losing precious seconds on their overall.
this is my concern with aerogels, as they break down, do they become problematic and cause things like silicosis? I wonder about it because I often experiment with different ways to make larger and larger sheets of aerogel for the home appliance and home insulation markets.
guilty of what?!? I know, I read the blogs and heard the arguments about what the jury found him guilty of something. When you can make anything a crime, everything can be a crime.
Remember that the next time you help out a friend and they tell you their password.
He withheld passwords. That's the long and short of it.
It blows my mind that the guy spent any time at all in jail for this, especially after the city lied about the access (they had access several days before he tuned over the passwords). It's worse when the city again lied, time and time again, in fact, in painting his actions and configurations as nefarious when they're all common practice. The sniffer thing, the modem stuff, the paging issue. Those lies the city told should have been a get out of jail free card for him by painting the city as the scumbags they are.
He did one thing wrong to his bosses, his bosses (via lawyer proxy, I assume) then turn around and lie in court, which is the real crime.
b) How often has a war against an ideal ever resulted in victory? Martyrs usually make ideals stronger. Are there more or less Taliban now than in 2001?
ww2, global war against general European fascism, German Nazism and Japanese Imperialism. Seems like it was pretty successful.
Cold War, global war against communism, brought down the Soviets after a while, though the jury is still out on the cost of this one.
BSG played this issue just right. They gleefully killed off main characters all the way through the series. They had the Cylons resurrected, but the humans didn't, and they made a point of scaring the shit out of the Cylons when that ability was lost. I thought it made great use of this plot device.
> It's crazy that the NYT spent so much money on a paywall. It makes no sense at all. They should have > funded 40 start ups to try to find a way to make more money from the news. Rather than trying to innovate > and find the future of news, they spend money trying to preserve their old business model.
QFT. They could have funded 40 start ups to try and find a way to make money (not necessarily on the news). This project should have cost around $50k and a week of development time.
This ludicrous overhead is the reason the NYT will cease to be a functioning business in the near future.
I'm caught. My son's science fair project tonight is about time travel, interestingly enough.
I may as well answer. $12. Each. The Yankees win the world series. Again.
It was hell being in the time machine in the rented storage locker for so long, but I slept through most of the waiting and, well, you know, for the rest.
Slashdot Mirror
not claiming a bug, just don't remember ever setting my posting preference to "CODE". though if that's a legitimate setting, why does it bug people that it's in use and why don't more people use it?
i changed it to plain text, because I don't like controversy.
nope, we're counting all reasons for rejections in those figures reported, we don't block by IP except on temporary basis or what's in well established RBLs, and listings in those RBLs are all temporary (no use of permanent RBLs like that one particular one who blackmails people into paying $50 to get their IP off the list after baiting senders with subscriptions)
Let's say you notice 10.10.10.0/24 has only ever sent junk, why not block the entire class C?
If a reliable reputation/feedback database says that mail originating in that subnet is reported as junk 100% of the time, why allow that subnet to waste your money?
that would require actually reading the article and comprehending what I read. people ask too much around here. Sheesh ;-)
sorry, just hit reply, and that's the font that came up after preview/submit. I'm not normally a LOOK AT ME!!! type of guy. Well, I am. Just in this case it was inadvertent.
long term, we've been allowing into the environment roughly the same volume of email per customer for 10 years. Some spam gets through, most does not, and there are few false positives. those that are labeled false positives are most often bulk mail that people mark as junk. So IMO, it's junk mail.
We use rules at the protocol level, DNS responses, RBLs (combined into one large RBL with miltiple return values), external reputation lists, internal dynamic reputation lists, rate limitations, and multiple feedback systems to provide this level of protection, that's before content filtering and personal white/black lists.
Just today, on the protocol layer, we're blocking 60% at banner (RBLs, bad DNS) , %14 of the remainder at HELO, %3.5 of the remainder Mail From (fake domain names) and finally a good chunk of what's left is blocked because it's destined to bad email addresses (which feeds back into the reputation lists).
Customer feedback helps stop those who are newly spewing spam, and since the feedback systems are widely distributed over many different email service providers, a massive spike at one translates into a blocked email at the others (whether by IP or content).
Better still, we do the same thing on the outbound side of things. If a customer catches a virus, they're cut off from email pretty fast and the feedback system is a very very tight loop internally.
But you are right, it's an ever escalating war, and if we could skip a few steps and jail (permanently, with broken hands) the spammers and bot coders, we wouldn't have to spend the money on the filtering and RBLs and feedback loops and hardware. We adjust the rules slowly over time, the feedback systems are maintained by the "trusted" customer, we're spending hundreds of thousands of dollars a year to protect against junk mail. I'm not certain of the math here, but an educated guess, this translates to around %5 of the cost to serve a user's mailbox. That's just operations staff time, and datacenter space for the extra hardware, the hardware itself, the subscription fees to the antispam service, wasted bandwidth etc.
you are correct, the missing data point is the volume of email considered "not spam". This line in the graph stayed the same over the range, or within a minor fraction of a percent of the same. it's the spam counts that have dropped since 10/2010. The customer base also represents a large number of domain names, hundreds of thousands of domain names. One of our largest customers has been offering email since 1995, with many accounts in their domain being around for over a decade. I think it's a pretty solid sample of email accounts.
my graphs show a steady decline in spam capture rates since October, 2010. we're measuring an average daily rate about 1/2 of this time last year. (millions of mail boxes, dozens of MX servers, decent antispam filtering) We're blocking around %91.2 of mail at the perimeter as opposed to %98.8 last year.
I work for an email service provider, we're catching many each day, most less than 500 emails at a time. I think about 1/2 of them are compromised PCs as they're using the same IP addresses the customers use, different HELO hostname and all that but they're still authenticating from the same place. That's the wild part. I watched a network sniff play out on screen, showed the authentication stuff, same user ID and password, different HELO hostname and headers, right along side another session where the user was sending legit email.
The other portion are clearly phished accounts, customer in Boise, connection from China for example.
The kicker is that we've had to turn off our internal reputation system based on the age of the email account. Used to be > 1 month old had higher limits than < 1 month old (for the love-em-then-leave-em accounts), but today, no one is trusted.
The only good thing is they seem to come in phases, where a particular campaign of the exact same email comes from dozens of accounts, for hours at a time, then switches to a new campaign later. Makes filters easier to manage.
If you're not paying for it, they can do whatever they want with it, how else are they supposed to make money off the customer for providing the service?
Don't like it, change your email address.
They'll wind up like Excite, for example. Too many freeloaders, not enough clicks on ads.
that's not trolling, that's the truth.
they get unexplained skin growths from crapping in their suits because they're too hard core take the time to find a port-a-potty and risk losing precious seconds on their overall.
eyeball goo changes somewhat over the years, Vitreous Humor it's called, and perhaps that change is what caused your direction finding to fade
East Blvd is named after someone, East Ave heads East. as West Ave heads West.
ouch, cringing just thinking about the last time I did that, and it was 25 years ago. my foot removed some of the crayon
what are the craters around the area from? nuke tests or conventional weapons tests or just holes they dug for some other reason?
this is my concern with aerogels, as they break down, do they become problematic and cause things like silicosis? I wonder about it because I often experiment with different ways to make larger and larger sheets of aerogel for the home appliance and home insulation markets.
we'll move deimos to become the anchor for the mars elevator...
guilty of what?!? I know, I read the blogs and heard the arguments about what the jury found him guilty of something. When you can make anything a crime, everything can be a crime.
Remember that the next time you help out a friend and they tell you their password.
He withheld passwords. That's the long and short of it.
It blows my mind that the guy spent any time at all in jail for this, especially after the city lied about the access (they had access several days before he tuned over the passwords). It's worse when the city again lied, time and time again, in fact, in painting his actions and configurations as nefarious when they're all common practice. The sniffer thing, the modem stuff, the paging issue. Those lies the city told should have been a get out of jail free card for him by painting the city as the scumbags they are.
He did one thing wrong to his bosses, his bosses (via lawyer proxy, I assume) then turn around and lie in court, which is the real crime.
fusion: in which case you check their atmosphere for extra helium and the aliens for an extra arm.
b) How often has a war against an ideal ever resulted in victory? Martyrs usually make ideals stronger. Are there more or less Taliban now than in 2001?
ww2, global war against general European fascism, German Nazism and Japanese Imperialism. Seems like it was pretty successful.
Cold War, global war against communism, brought down the Soviets after a while, though the jury is still out on the cost of this one.
BSG played this issue just right. They gleefully killed off main characters all the way through the series. They had the Cylons resurrected, but the humans didn't, and they made a point of scaring the shit out of the Cylons when that ability was lost. I thought it made great use of this plot device.
rishathra?
> It's crazy that the NYT spent so much money on a paywall. It makes no sense at all. They should have
> funded 40 start ups to try to find a way to make more money from the news. Rather than trying to innovate
> and find the future of news, they spend money trying to preserve their old business model.
QFT. They could have funded 40 start ups to try and find a way to make money (not necessarily on the news). This project should have cost around $50k and a week of development time.
This ludicrous overhead is the reason the NYT will cease to be a functioning business in the near future.
I'm caught. My son's science fair project tonight is about time travel, interestingly enough.
I may as well answer. $12. Each. The Yankees win the world series. Again.
It was hell being in the time machine in the rented storage locker for so long, but I slept through most of the waiting and, well, you know, for the rest.