On Wednesday 24 July 2019 10:25:51 Damian wrote: > Hi, > > > Would you mind to test if amavisd-new work with DBD::MariaDB? As > > after two years people still reports that amavisd-new has problems with > > DBD::mysql, it would be probably a good idea to look if DBD::MariaDB > > driver is not more suitable for end users. > > is DBD::MariaDB intended to be and remain a drop-in replacement for > DBD::mysql or will the feature sets diverge eventually (however that > would look like)?
Hi! I want to have DBD::MariaDB implemented according to DBI API (as documented in https://metacpan.org/pod/DBI) and to be compatible with other DBI drivers, like DBD::Pg or DBD::SQLite. So if there is some bug in DBD::MariaDB which says that driver behavior different as should be according to DBI API then DBD::MariaDB driver is intended to have fixed this bug, as opposite to DBD::mysql which are not going to fix such bugs. So DBD::MariaDB may diverge, but because of fixing bugs. It already happen e.g. for those floating point values (DBD::mysql returns sometimes zeros and amavis does not work; DBD::MariaDB returns correct values) or for Unicode data (DBD::MariaDB always returns and accept them correctly). > The former would be most convenient, as DBD::MariaDB > could be proposed as an officially supported alternative for amavis > without code change. If amavis is written in way that does not use private driver functions, and only public DBI API then it should work fine.
