<=====================================8<=====================================>
Posted By: Roumen (Roumen) on 'Koudink'
Date: Tue 13.9.2005 21:02.57
Title: Re: CZ.NIC
Kam jdou ty penize, nevim, ale silne pochybuji, ze za vsechny existujici
domeny v soucasne dobe jejich vlastnici plati, takze bych rekl, ze ta
ohromujici castka bude o neco mensi (moc ale nevidim do vztahu
NIC-registratori, takze se mozna pletu).
Mozna jo. Nezaplacena domena je mesic v ochranny lhute a pak jde do kytek.
Kdyz nic jinyho, muze ti ji nekdo slohnout...
Roumen
For more information about me use:http://www.roumen.cz/
<=====================================8<=====================================>
Posted By: Xofon (==(O)==@@@@=@@@@===) on 'Koudink'
Date: Tue 13.9.2005 23:50.05
Title: online deskove hry
Koukal jsem, jestli ve starsim debianu neni nejaky balik na snadne hrani
sitovych piskvorku. Nenasel jsem nic, co by mi vyhovovalo, tak jsem to
vyresil nalogovanim nas obou na stejny pocitac a takovymto programem v bashi.
while [ -f .pis.swp ]; do sleep 1; done; vi pis
:-)
Ma to i integrovany chat.
Pokud to je vseobecne znama finta, tak se omlouvam. Ale stejne, az to
nekdy budu potrebovat, tak to najdu tady.
Xof
--
ebbs: /home/bbs/xofon/.signature: No such file or directory
<=====================================8<=====================================>
Posted By: Harvester (12.110.110.204) on 'Koudink'
Date: Thu 29.9.2005 8:16.26
Title: Bayes Antispam filter
Pouzivam zmineny filter v postovnim klientu The Bat! ve Windows. Je fajn,
akorat posledni dobou se mi zda, ze je dost neucinny. Resp. premyslim, jestli
spameri nasli zpusob jak koncipovat text emailu tak, aby filtrem prosly.
Na zaklade tohoto chovani premyslim, co s tim:
a) snizit skore pro spam
b) co se stane, kdyz zname spam mejly projedu filtrem manualne jeste jednou,
pripadne nekolikrat? Bude na ne filtr citlivejsi nebo se nestane nic?
_________
Harvester
<=====================================8<=====================================>
Posted By: Covex (mi'lius) on 'Koudink'
Date: Thu 29.9.2005 14:51.10
Title: Re: Bayes Antispam filter
Pouzivam zmineny filter v postovnim klientu The Bat! ve Windows. Je fajn,
To je takovy bayny vselek na spam.
akorat posledni dobou se mi zda, ze je dost neucinny. Resp. premyslim,
jestli
spameri nasli zpusob jak koncipovat text emailu tak, aby filtrem prosly.
Na zaklade tohoto chovani premyslim, co s tim:
a) snizit skore pro spam
Budes mic vit hamu.
b) co se stane, kdyz zname spam mejly projedu filtrem manualne jeste jednou,
Ne vzdy znamena vice lepe. Tak nejak bych rekl, ze podstatna je tez kvalita
stravy. Pokud se spameri uci psat kulturne moc s tim nenadelas, casem stejne
vsichni prejdem na whitelisting.:)
pripadne nekolikrat? Bude na ne filtr citlivejsi nebo se nestane nic?
_________
Harvester
Covex
<=====================================8<=====================================>
Posted By: Tuttle (back to the konami times) on 'Koudink'
Date: Thu 29.9.2005 15:14.49
Title: Re: Bayes Antispam filter
casem stejne vsichni prejdem na whitelisting.:)
Tot velmi odvazne tvrzeni. Nesouhlasim s nim, ale je odvazne. :-)
T.
<=====================================8<=====================================>
Posted By: Covex (mi'lius) on 'Koudink'
Date: Thu 29.9.2005 16:01.24
Title: Re: Bayes Antispam filter
Ne vzdy znamena vice lepe. Tak nejak bych rekl, ze podstatna je tez kvalita
stravy.
Training
Training is the process of feeding known ham and spam into the tokenizer to
generate the probabilities that the classifier then uses to generate it's
decisions. A variety of approaches to training have been tried - things that
we found:
* the system generates remarkably good results on very few trained
messages, although obviously, the more the better.
* training on an extremely large number of messages may start to produce
slightly poorer results (once you're training on thousands and thousands of
messages).
* imbalance between the number of ham and spam trained can cause
problems, and worse problems the greater the imbalance; the system works best
when trained on an approximately equal number of ham and spam.
* It's really important that you're careful about selecting your training
sets to be from the same sources, where-ever possible. Several people got
caught out by training on current spam, and old ham. The system very quickly
picked up that anything with a 'Date:' field with a year of 2001 was ham. If
you can't get ham and spam from the same source, make sure the tokenizer
isn't going to pick up on the non-clues.
* "mistake-based training" - where you only train on messages that the
system got wrong - results in the system taking a lot longer to get really
good at what it does. You're better off feeding it the things it gets right,
as well as the things it gets wrong.
Covex
Covex
<=====================================8<=====================================>
Posted By: CePal (ICQ # 60033556) on 'Koudink'
Date: Thu 29.9.2005 21:17.02
Title: Re: Bayes Antispam filter
casem stejne vsichni prejdem na whitelisting.:)
Tot velmi odvazne tvrzeni. Nesouhlasim s nim, ale je odvazne. :-)
T.
Ja s nim souhlasim... vsechno k tomu smeruje...
CePal
<=====================================8<=====================================>
Posted By: Harvester (12.110.110.204) on 'Koudink'
Date: Fri 30.9.2005 10:55.02
Title: Re: Bayes Antispam filter
a) snizit skore pro spam
Budes mic vit hamu.
No, myslel jsem snizit jako oznacovat jako spam vetsi procento mejlu.
b) co se stane, kdyz zname spam mejly projedu filtrem manualne jeste
jednou,
Ne vzdy znamena vice lepe. Tak nejak bych rekl, ze podstatna je tez kvalita
stravy. Pokud se spameri uci psat kulturne moc s tim nenadelas, casem stejne
vsichni prejdem na whitelisting.:)
Uz aby. Kdysi jsem psal do Ritlabs, jestli nechteji implementovat funkci ala
whitelisting, ktera tam sice je, ale "jen" whitelisting. Ma predstava je
takova, ze kdo neni ve whitelistu, od toho se stahne jen hlavicka. Nemam
prehled o ostatnich klientech treba i pod Linuxem, jestli uz to nektere
delaji, ale podle me je to presne to, co by nejvic pomohlo. I sebelepsi
antispam filter na stanici ten mejl jen tridi, ale stahne ho vzdycky. Ted na
broadbandu je mi to celkem jedno, byly vsak i doby dialupu, kde by to hralo
dost podstatnou roli.
pripadne nekolikrat? Bude na ne filtr citlivejsi nebo se nestane nic?
_________
Harvester
Covex
_________
Harvester
<=====================================8<=====================================>
Posted By: pharook (dee do de de) on 'Koudink'
Date: Fri 30.9.2005 19:01.35
Title: Re: Bayes Antispam filter
Uz aby. Kdysi jsem psal do Ritlabs, jestli nechteji implementovat funkci ala
whitelisting, ktera tam sice je, ale "jen" whitelisting. Ma predstava je
takova, ze kdo neni ve whitelistu, od toho se stahne jen hlavicka. Nemam
prehled o ostatnich klientech treba i pod Linuxem, jestli uz to nektere
delaji, ale podle me je to presne to, co by nejvic pomohlo. I sebelepsi
antispam filter na stanici ten mejl jen tridi, ale stahne ho vzdycky. Ted na
broadbandu je mi to celkem jedno, byly vsak i doby dialupu, kde by to hralo
dost podstatnou roli.
Harvester
Pragmaticni spravci implementuji rozumnou cast antispamove a antivirove
ochrany primo na serveru; nekdy i celou. Pragmaticti uzivatele si vybiraji
takove servery. :)
____________________________________________________________________pharook_
"Mesic je dulezitejsi nez Slunce", reklo dite. "Protoze sviti, kdyz je tma."
<=====================================8<=====================================>
Posted By: pharook (dee do de de) on 'Koudink'
Date: Fri 30.9.2005 19:05.11
Title: Re: Bayes Antispam filter
Ne vzdy znamena vice lepe. Tak nejak bych rekl, ze podstatna je tez kvalita
stravy. Pokud se spameri uci psat kulturne moc s tim nenadelas, casem stejne
vsichni prejdem na whitelisting.:)
Covex
Ja doufam spis v podpisy. Obavam se, ze jenom doufam. :)
____________________________________________________________________pharook_
"Mesic je dulezitejsi nez Slunce", reklo dite. "Protoze sviti, kdyz je tma."
|
