Proposal Seeks To Replace MySQL With MariaDB By Default For Ubuntu 25.04

[calica]

My limited exposure to MySQL (and by extension MariaDb) was enough to convince me anyone should be using PostgreSQL or SQLite unless they're already invested in MySQL, though, so I've never had a horse in the race between MySQL and MariaDb.

That was the case with Postgres95 and mSQL almost 30 years ago.

[J.King]

maria/mysql aren't drop in compatible.

for example from airflow documentation:



while i find that statement to be a bit FUDish, it's probably true that maria went too far by being a hard fork rather than being a mysql++, but I think the bigger issue was that postgresql has most of the mindshare for developers looking for an opensource database. I personally like postgresql, but i find it very frustrating to administer.

My own software supports MySQL but not MariaDb. Both databases support common table expressions (and MariaDb did so first), but MySQL's implementation is comprehensive and correct whereas MariaDb's has multiple limitations/bugs. There's been an open bug report about it for over five years with no resolution in sight:

https://jira.mariadb.org/browse/MDEV-18511

My limited exposure to MySQL (and by extension MariaDb) was enough to convince me anyone should be using PostgreSQL or SQLite unless they're already invested in MySQL, though, so I've never had a horse in the race between MySQL and MariaDb.

[ahrs]

I think masking mysql packages as actual mariadb packages was a terrible mistake by distros. They should have kept mysql and package mariadb separately, let the end user decide and stats would show true end user preference rather than masking it and pushing mariadb down our throats.. If I want to install mysql, give me mysql god da***!

The point of doing that was to provide a drop-in replacement and easy upgrade path but apparently Fedora no longer considers them to be equal anyway:

https://lists.ubuntu.com/archives/ubuntu-devel/2024-September/043131.html

[Espionage724]

That's completely irrelevant. The subject wasn't about MariaDB. It was about a project with broken build infrastructure that is intended to build with an embedded copy of a dependency that happens to pick up a system copy of a dependency when it should not. That's the fault of TrinityCore and how their build infrastructure has been written. Fix the header paths so they're in the right order as intended.

FreeBSD is packaging it how it should be. They're not doing anything wrong.

The problem was not MariaDB.

Ah, that makes the counter make more sense now! And I learned something

[brad0]

You're not convincing me MariaDB is a solve-all solution when the project I used for years longer than I even knew MariaDB existed worked perfect... for years with MySQL

But fine; semantics demands I correct that to not put direct-blame of MariaDB, and I realize that FreeBSD could be packaging it specifically that way for a reason (not feeling like re-wording it myself, you do it ). Point was, MariaDB was a problem for me 2 times, so why would I want to use it still let alone vouch for it? And what are you defending exactly to come at me with a counter?

That's completely irrelevant. The subject wasn't about MariaDB. It was about a project with broken build infrastructure that is intended to build with an embedded copy of a dependency that happens to pick up a system copy of a dependency when it should not. That's the fault of TrinityCore and how their build infrastructure has been written. Fix the header paths so they're in the right order as intended.

FreeBSD is packaging it how it should be. They're not doing anything wrong.

The problem was not MariaDB.

[Espionage724]

MariaDB didn't "cause" anything. The TrinityCore build infrastructure is broken and needs fixing.

You're not convincing me MariaDB is a solve-all solution when the project I used for years longer than I even knew MariaDB existed worked perfect... for years with MySQL

But fine; semantics demands I correct that to not put direct-blame of MariaDB, and I realize that FreeBSD could be packaging it specifically that way for a reason (not feeling like re-wording it myself, you do it ). Point was, MariaDB was a problem for me 2 times, so why would I want to use it still let alone vouch for it? And what are you defending exactly to come at me with a counter?

[brad0]

On FreeBSD 14.1, I ran into MariaDB causing that project to fail (ticket) because MariaDB had fmt as a dependency which caused it to get used instead of that project's fmt. MySQL on FreeBSD didn't have the fmt dep, it installed, and that project compiled fine!

MariaDB didn't "cause" anything. The TrinityCore build infrastructure is broken and needs fixing.

[Old Nobody]

Sometimes Canonical decisions are very puzzling and this is one of them. I wonder why they waited this long. For comparison, Fedora did this over 10 years ago and RHEL followed soon after. openSUSE and SUSE did this as well around the same time. If they are consistently conservative like Debian that's one one thing but they recently announced they are going to not use LTS kernels and instead ship more bleeding edge ones. So I don't follow the logic here.

Debian defaults to MariaDB for more than 7 years.

[Espionage724]

Well then you could run Ubuntu Core on something like a Raspberry Pi, because Ubuntu Core can only install Snap packages, not .deb packages.

Yeah I'd just install normal Ubuntu (or Server) and set it up for real But I guess that's a cool way that it could be used!

[uid313]

While I'm sure that could be practical in some way, that's (imo) beyond gross to ever think of existing let alone using

Well then you could run Ubuntu Core on something like a Raspberry Pi, because Ubuntu Core can only install Snap packages, not .deb packages.