Friend, you need to take a look at the specs on CRM114 at crm114.sourceforge.net. While the interface and initial setup are fairly painful for people who don't build their own email setups, various folks are publishing that they get over 99.9% correct detection of both spam and non-spam. That's far better than any other single filter out there.
Re:Simple
on
The 3Com Saga
·
· Score: 2, Interesting
Hmm. What I read from your description of 3com onboard being wildly inferior to 3com on a network card matches my experience.
3com onboard is so *bad* it makes 3com on a card look wildly better by comparison.
But if you need wildly better performance on a card, invest in a Kingston 4-port card: those still use the old DEC Tulip chipsets or a newer chipset with the DEC scraped off and a new manufacturer. These are not "tulip compatible", they're actually Tulip chipsets, and goodness do they work well under high load. Interrupt coalescing, here I come!
3com has, in my experience with thousands of Os installations, sucked badly.
They are the *only* network card vendor whose 10Base2 cards traditionally shorted the network ground to system ground after 2 years of use. Try dealing with that sort of nonsense in a low-signal analog measurement system. It's a nightmare to track down.
Then various of their much vaunted 100BaseT chipsets often fail to correctly reset DHCP unless you do a cold reboot, their wildly overpriced GigE cards change chipsets and change names for their drivers while keeping the same name on the package with a tiny, tiny little suffix as the only hint, they refuse to actually list the chipsets used on their specsheets, they wildly vary and overlap the names of the drivers in their Windows installations so much as to make it impossible to tell which card you're actually installing drivers for, their router language was designed by drunken monkeys using a spam random word generator, etc., etc.
As you might guess, I'm *not* happy with 3com, and haven't been for a while. If you're really looking for 100BaseT cards, spend the money and actually find an old DEC Tulip chipset. Not "tulip compatible", I mean the real honest-to-ghod old Tulip chipset, which was incredibly robust and has never been beaten for network performance under load.
But 3Com? Don't make me wince.
Unfortunately, it don't work that way. Most of us open source developers are toolmakers, creating tools to fit a need. The closed source world prevents us from sharing the tools, and incredibly limits our creativity by forcing us to all buy our own toolkits: so the student programmer can't use, or tweak, the C compiler unless they shell out $5000 for a source code license, for example.
So tools simply don't get created, or their bugs are concealed and cannot be repaired except by the Authorized Vendor (and their often over-worked support staff).
In the open source world, the payment we make is for my salary to go tweak the tools to fit our site's particular needs. I get paid alright, and they get tools customized to their needs, and the tools get revised *quickly* to fit new requirements. And the overall quality is much higher because gifted people poke their noses into the tent and share their tools from other open source projects for me to work with (such as gcc, the OpenSSL libraries, Perl, Python, etc., etc.)
Having the right tools available for the cost of your engineers' open source development time is a very, very good trade in most cases.
China has actually been entire blocked before. Quite some years ago, when their primary uplink feed was a pair of ISDN lines to allow their censors to minitor all traffic, they lost their link from their upstream when over 50% of all traffic was spam.
Their bandwidth has gotten larger since then. But really, a lot of the problem is the APNIC DNS authority, who allow fraudulent DNS sites and don't seem willing to implement an anti-spam providers. Couple that with with *no* broad standards for how to block relaying and forgery of email addresses, and you have a disaster.
Folks, you need to hop over to http://spf.pobox.com and see what's going on in the new SMTP RFC's, designed to allow control over what mail is sent in your name. It's fascinating fair, and may help one hell of a lot against the viral traffic and the "forged from your domain to get past your whitelist" spam.
Slashdot Mirror
Friend, you need to take a look at the specs on CRM114 at crm114.sourceforge.net. While the interface and initial setup are fairly painful for people who don't build their own email setups, various folks are publishing that they get over 99.9% correct detection of both spam and non-spam. That's far better than any other single filter out there.
Hmm. What I read from your description of 3com onboard being wildly inferior to 3com on a network card matches my experience. 3com onboard is so *bad* it makes 3com on a card look wildly better by comparison. But if you need wildly better performance on a card, invest in a Kingston 4-port card: those still use the old DEC Tulip chipsets or a newer chipset with the DEC scraped off and a new manufacturer. These are not "tulip compatible", they're actually Tulip chipsets, and goodness do they work well under high load. Interrupt coalescing, here I come!
3com has, in my experience with thousands of Os installations, sucked badly. They are the *only* network card vendor whose 10Base2 cards traditionally shorted the network ground to system ground after 2 years of use. Try dealing with that sort of nonsense in a low-signal analog measurement system. It's a nightmare to track down. Then various of their much vaunted 100BaseT chipsets often fail to correctly reset DHCP unless you do a cold reboot, their wildly overpriced GigE cards change chipsets and change names for their drivers while keeping the same name on the package with a tiny, tiny little suffix as the only hint, they refuse to actually list the chipsets used on their specsheets, they wildly vary and overlap the names of the drivers in their Windows installations so much as to make it impossible to tell which card you're actually installing drivers for, their router language was designed by drunken monkeys using a spam random word generator, etc., etc. As you might guess, I'm *not* happy with 3com, and haven't been for a while. If you're really looking for 100BaseT cards, spend the money and actually find an old DEC Tulip chipset. Not "tulip compatible", I mean the real honest-to-ghod old Tulip chipset, which was incredibly robust and has never been beaten for network performance under load. But 3Com? Don't make me wince.
Unfortunately, it don't work that way. Most of us open source developers are toolmakers, creating tools to fit a need. The closed source world prevents us from sharing the tools, and incredibly limits our creativity by forcing us to all buy our own toolkits: so the student programmer can't use, or tweak, the C compiler unless they shell out $5000 for a source code license, for example. So tools simply don't get created, or their bugs are concealed and cannot be repaired except by the Authorized Vendor (and their often over-worked support staff). In the open source world, the payment we make is for my salary to go tweak the tools to fit our site's particular needs. I get paid alright, and they get tools customized to their needs, and the tools get revised *quickly* to fit new requirements. And the overall quality is much higher because gifted people poke their noses into the tent and share their tools from other open source projects for me to work with (such as gcc, the OpenSSL libraries, Perl, Python, etc., etc.) Having the right tools available for the cost of your engineers' open source development time is a very, very good trade in most cases.
China has actually been entire blocked before. Quite some years ago, when their primary uplink feed was a pair of ISDN lines to allow their censors to minitor all traffic, they lost their link from their upstream when over 50% of all traffic was spam. Their bandwidth has gotten larger since then. But really, a lot of the problem is the APNIC DNS authority, who allow fraudulent DNS sites and don't seem willing to implement an anti-spam providers. Couple that with with *no* broad standards for how to block relaying and forgery of email addresses, and you have a disaster. Folks, you need to hop over to http://spf.pobox.com and see what's going on in the new SMTP RFC's, designed to allow control over what mail is sent in your name. It's fascinating fair, and may help one hell of a lot against the viral traffic and the "forged from your domain to get past your whitelist" spam.