Re: rspamd and pyzor

03.05.2022 16:48, Stuart Henderson пишет:
> On 2022-05-03, kasak <kasak@kasakoff.net> wrote:
>> rspamd manual assume, that we should use this construction:
>>
>> ExecStart=/bin/sh -c '/usr/bin/razor-check && /usr/bin/echo -n "spam" ||
>> /usr/bin/echo -n "ham"'
>>
>> The razor-check manual confirm this: "razor-check" terminates with exit
>> value 0 if the signature for the mail is catalogued on the server (spam)
>> or 1 if the mail is not catalogued by the server (not a spam).
>>
>> I don't like this construction, and can't even imagine that we can do
>> the same with inetd.
>> Maybe i should put all this to some script and call it from inetd, but
>> i'm not sure it this a good idea or not.
> it may work direct in inetd.conf, but yes that would be neater.
> I don't think there's any disadvantage to running it in a shell script
> rather than as a "sh -c" construct.
>
> btw keep an eye on how it performs, I haven't used pyzor/razor in a
> fairly long time but have used dcc more recently, I found it fairly easy
> for some bulk/automated non-spam to get caught up by the checks.
> I stopped using it because of this, the biggest problem was mail where
> there were a couple of other spam signs (but not enough to trigger
> detection by themselves). at least, you really don't want to learn
> symbols for Bayesian detection or feed neural network detection from
> those mails.
>
Thanks for the warning, I'll monitor it!