SeLLeRoNe
Super Moderator
I absolutely agree that it's better to use the OS provided ones, unless for some reason a newer version is required and available in CB.
Regarding the "use Rspamd RPM" I don't agree to much, I prefer software compiled on the box as for all the software (or almost all) in CB, otherwise why not simply do the same for Apache, Nginx and so on.
The ability to build from source will always allow to enable/add feature that the RPM don't come immediatly with.
As, for example, in exim not using the pre-compiled package allow me to enable experimental features as Generate DMARC Report which wouldn't be available otherwise.
If you have a test box I would still suggest to run a test doing the symlink.
I did some test on CentOS and it is working fine, of course I used a testdomain so I don't receive spam at all, that's why I am seriously considering to install it on a Production Server for a proper test, also keeping in mind that Rspamd can read Spamassassin configurations and rules, which make very easy to switch over without losing rules that I (or my customers) configured.
The only bit I am a little concerned about (but I need to do some reading about this) is the "Learning" as I use a lot sa-learn to make Spamassassin understand, on a per-email account basis, what each single user consider Spam and what Ham and I don't know if also that is "ported" into Rspamd and how to work with it in Rspamd to to the same based on Inbox and Spam folders.
Regarding the "use Rspamd RPM" I don't agree to much, I prefer software compiled on the box as for all the software (or almost all) in CB, otherwise why not simply do the same for Apache, Nginx and so on.
The ability to build from source will always allow to enable/add feature that the RPM don't come immediatly with.
As, for example, in exim not using the pre-compiled package allow me to enable experimental features as Generate DMARC Report which wouldn't be available otherwise.
If you have a test box I would still suggest to run a test doing the symlink.
I did some test on CentOS and it is working fine, of course I used a testdomain so I don't receive spam at all, that's why I am seriously considering to install it on a Production Server for a proper test, also keeping in mind that Rspamd can read Spamassassin configurations and rules, which make very easy to switch over without losing rules that I (or my customers) configured.
The only bit I am a little concerned about (but I need to do some reading about this) is the "Learning" as I use a lot sa-learn to make Spamassassin understand, on a per-email account basis, what each single user consider Spam and what Ham and I don't know if also that is "ported" into Rspamd and how to work with it in Rspamd to to the same based on Inbox and Spam folders.