Fedora 19 Nixing MySQL in Favor of MariaDB

An anonymous reader writes "Red Hat developers are planning to replace MySQL with MariaDB in Fedora 19. For the next Fedora update, the MariaDB fork would replace MySQL and the official MySQL package would be discontinued after some time. The reasoning for this move is the uncertainty about Oracle's support of MySQL as an open-source project and moves to make the database more closed." Update: 01/22 13:47 GMT by T : Note: "Nixing" may be a bit strong; this move has been proposed, but is not yet officially decided.

Comments

See this comparison. Wikipedia is moving, too.

[phaunt]

Here is a comparison of MariaDB vs MySQL.
Probably most important to Fedora is this:

Truly Open Source

• All code in MariaDB is released under GPL, LPGL or BSD. MariaDB does not have closed source modules like the one you can find in MySQL enterprise edition. In fact, all the closed source features in MySQL 5.5 enterprise edition are found in the MariaDB open source version.
• MariaDB includes test cases for all fixed bugs. Oracle doesn't provide test cases for new bugs fixed in MySQL 5.5.
• All bugs and development plans are public.
• MariaDB is developed by the community in true open source spirit.

Wikipedia, too, is moving from MySQL to MariaDB.

Re:See this comparison. Wikipedia is moving, too.

[David Gerard]

Debian is planning to do the same (the thread containing approval from relevant people at Ubuntu too), for much the same reasons.

Re:See this comparison. Wikipedia is moving, too.

[cshark]

Two things that amaze me about this

1. That Debian isn't the first to do it. They're usually on top of this.

2. That it's taken this long to happen. We knew something like this would happen in the Linux community when Oracle took over MySQL. And there were some promising forking efforts that started immediately. Then, silance... for years. Now it's happening all at once, which is good.

Does anyone know if Maria is a MySQL fork? Or if it's totally new code? My cursory scan of the website hasn't revealed anything to this end.

Re:See this comparison. Wikipedia is moving, too.

[hairyfeet]

Ya know, you gotta give the little twerp that sold MySQL credit, he not only gets a big fat check but he then forks what he sold and gets all his customers back! Smart as hell, I'll give him that, as nowhere else but FOSS could you sell something and keep it at the exact same time

Re:See this comparison. Wikipedia is moving, too.

[hublan]

Um. He sold it to Sun. It wasn't until Sun got sold to Oracle -- a move which he fought against -- that the fork happened.

Postgresql

[Anonymous Coward]

Just use it.

Re:Postgresql

[egr]

Off-topic. Fedora already has PostgreSQL in its repositories. The point of the move is to replace MySQL, not to have some killer database.

Re:Postgresql

[petermgreen]

Unfortunately it's not that simple.

For example I have an application that uses a case insensitive collation. Afaict postgresql does not support this. There are ways to implement the same functionality (create an index on the uppercased version of the columns value) but it would mean changing every query that hits the columns in question.

For new stuff I will definately be choosing postgresql over mysql though.

P.S. does anyone know of a tool that can be used to design postgresql database schemas and export create/update scripts? (like mysql workbench does for mysql)

Re:Postgresql

[vlm]

P.S. does anyone know of a tool that can be used to design postgresql database schemas and export create/update scripts? (like mysql workbench does for mysql)

Could it be as simple as keep using mysql workbench (which I've never used) and pipe it thru SQLfairy aka SQL::Translator (a sourceforge project) to convert from mysql to postgresql?

I have a simple automated system that mysqldumps all my schema, then shoves them thru sqlfairy to convert to DOT (you'd be converting to postgresql, which I know sqlfairy claims to be able to do), then shoves the DOTs thru graphviz to convert to png diagrams, then creates a simple webpage to link to each db diagram. Easy and fast and fun and automatic. Also pretty boring unless you properly define your foreign keys.

Over the decades I've found its rather hard to machine translate, expect plenty of hand editing fine tuning unless its truly a miracle product and the format specs on both sides are incredibly well defined, or are very simple..

Re:Postgresql

[bad-badtz-maru]

Set your LC_COLLATE environmental variable on the PG server (and any client machines). PG probably isn't finding a setting, so it's defaulting to "C". If you switch it to something like en_US, collation will be case-insensitive. You may have to reindex after making the change.

We've begun the move away from MySQL also

[C_Kode]

We've begun to move away from MySQL offical release also. Although we went with Percona rather than MariaDB.

Re:We've begun the move away from MySQL also

[Anonymous Coward]

They are fairly comparable in features and performance in most cases while both are arguably superior to MySQL. MariaDB uses code from Percona (XtraDB) and tends to be more bleeding edge while Percona is more conservative and stable. MariaDB is a community effort whereas Percona is an actual company who you can call (pay) for support if needed (they will support any flavor of MySQL but Percona is obviously their area of expertise).They are both good products in their own way and give back a lot to the community. .02

Feature/Perfornace Comparisons:
http://stackoverflow.com/questions/12671634/mariadb-vs-drizzle-vs-percona-sever-vs-mysql
http://vbtechsupport.com/657/
http://vbtechsupport.com/606/

Yes.

[Anonymous Coward]

The point of the move is to replace MySQL, not to have some killer database.

And for those of us who are tied to MySQL, it's nice to have an alternative now without the hassle of moving to a completely different DBM.

Migrating

[mwvdlee]

How easy is it to migrate from MySQL to MariaDB?
Is it truely drop-in replacement as in "you can develop to MySQL, then run MariaDB in production without worrying"?
Does it require converting current tables? Will it take a 10GB database all day to convert or will MariaDB just use the raw MySQL data files automagically?

Re:Migrating

[hholzgra]

> Is it truely drop-in replacement as in "you can develop to MySQL, then run MariaDB in production without worrying"?

yes, unless you use some of the non-GPL extra features like e.g. authentication plugins or pool-of-threads. For these MariaDB has GPL replacements but the implementation and configuration may differ ...

> Does it require converting current tables?

Data format of MyISAM and InnoDB tables is the same, so "no" in general. mysql system database may differ a bit, but nothing the mysql_upgrade tool can't fix, and you'll have the same issues when develop against an older MySQL version and deploying to a newer one ...

> Will it take a 10GB database all day to convert or will MariaDB just use the raw MySQL data files automagically?

It will use existing raw files just fine. mysql_upgrade may take a few minutes max., but not all day ... (unless you're migrating from an older MySQL version and mysql_upgrade needs to recreate some indexes ... but that would happen when upgrading to a more current MySQL release, too, and wouldn't be MariaDB specific

The only point where it isn't a simple "try and revert if you don't like it" drop in replacement is if mysql_upgrade changed mysql.* system tables and you want to roll back to regular MySQL ... but then again this is also the case when trying to upgrade to a more recent MySQL release and then deciding to roll back to a previous older one again ... so you should always have a backup to restore the original system tables from ... but you'd do a full backup before any version migrations anyway, wouldn't you?

Re:Migrating

[TheBlackMan]

I have recently migrated 2 of my servers from MySQL to MariaDB with about 1,5mil unique users. The only problems i had were some configuration (my.cnf) changes. Except that, none at all.

Re:Migrating

[deKernel]

Uhm these are all very good questions, but here is a suggestion: verify this for yourself!

Ha Ha

[EETech1]

That is one thing I always loved about Open Source Software, you can't hold it hostage. If something happens to threaten the project, its fork you, and bye bye!

The users and developers take what was there (and theirs) pack up and resume life a usual somewhere else, and give the finger to all the Larrys that come along and try to (ab)use them as part of their own personal plot.

If you take care of them they will stay, and you might even see your user and developer base grow, but try and take advantage of them and you will soon be left with nothing. They take the short term pain of starting over instead of the long term pain of taking orders from someone who does not have their best interests in mind.

Cheers!

Re:Ha Ha

[Kardos]

The lesson here is: Don't let Oracle touch anything open source, or you're going to be forking in a year or two

Re:Ha Ha

[cheesybagel]

The problem with Oracle is they always do it like this. First they acquire a major competitor in a major cash transaction then they milk it for a couple of years firing most of the development team and selling licenses while they can. Then they cease support altogether. They are not interested in further developing anything they acquire. It is all slash and burn. Their main expertise was never development it is sales and support.

Update: 01/22 14:01 GMT by LarryEllison:

[LulzAndOrder]

Larry Ellison responded with the comment, "you are all spelling it wrong, I renamed it MY sql some time ago."

Re:Too bad SQL is not like W3C standards

[Billly Gates]

Every vendor wants their own proprietary extenstions, and even fight each other on who gets the most proprietary lockin, apis, and extensions on top of their own SQL to maximize pain leaving their shit ecosystem. ADO.NET, ODBC, and other things where you can't just use sql, but another vendor $$$ framework that locks you in further and add dozens of lines of ugly code and before you add another proprietary layer of Vendor X SQL inside it.

Then gee you can't leave it. Larry grins and then raises the cost knowing you are hostage etc (dramatization here).

HTML and CSS was that way in the 1990s as well. We bash IE 6 here but Netscape didn't follow standards either. Today that is fixed.Why can't we do that with data access?

What will it take for all of them to work together like ansi ascii text, C++, or HTML?

I admit Oracle and SQL Server will be a bitch to get rid of in the enterprise due to these sneaky moves, but slowly change will come when PostgresSQL, MariaDB, and others who want to play nice come up with one standard, one way to do things, and get oustracized here on slashdot as crap if they do not follow the spec? This is 2013 and these proprietary games are oldschool before the internet when corps were in a pissing match on who could make the most crappiest proprietary system out there so customers could never leave... cough unix.

Database? As part of an operating system?

[sunderland56]

Why is *any* database a part of an OS? This isn't required functionality. It isn't even day-to-day useful as a browser, or a word processor.

After installing the base OS, a distro can offer to optionally install packages - such as a database - but I don't see why that choice should be limited to just one example. Make both MySQL and MariaDB available, and any others you want.

Re:Database? As part of an operating system?

[nine-times]

Well I'm sure part of the issue here isn't whether it's technically "part of the OS", but rather a question of the repositories. Since Linux distros these days generally include a package manager linked to a set of official/supported repositories, anything in those repositories becomes, in a sense, part of the operating system that they're distributing.

Fedora has to decide what they'll include in their repositories, what they'll treat as "supported", and what they'll list as "deprecated". Sure, you can go along afterwards and install all kinds of other things from source, but most people will use the repositories if they can, so what Fedora decides to include is an important issue.

slashdotted?

[luis_a_espinal]

Tehe, mariadb.org no workie.

Who trusts Monty Widenius?

[jjohnson]

MariaDB is Monty Widenius' fork after leaving Oracle. MySQL is in Oracle's hands in the first place because Widenius sold MySQL to Sun. The man responsible for MySQL is also responsible for the fragmentation of the community into a bunch of forks with growing incompatibility problems.

Re:Who trusts Monty Widenius?

[jjohnson]

Besides MySQL, MariaDB and Percona, there's Dribble that's actually a viable fork. One's official, one's competing by the original founder, one's got all the experts at clustering, and one's just kind of there as well. So now, instead of one weak-ass database, you have four that are mostly the same but not identical, which will bite you in the ass at some unspecified point in the future.

Back when it was just MySQL, it was a reasonable choice because it had known strengths and weaknesses. Now you've got shitty fragmentation. If you're starting a new project, which fork do you go with?

Easy answer: Postgres.

Re:Percona

[geminidomino]

So the developers of MariaDB took all their experience and knowledge that they obtained at Oracle while working on MySQL and created a direct competitor product to their prior company's product?

Yep, and he was SUCH a sneaky, underhanded shit that he left the company a whole YEAR before Oracle even knew they had it.

Derp.

Re:Percona

[alexgieg]

So the developers of MariaDB took all their experience and knowledge that they obtained at Oracle while working on MySQL

No. Oracle purchased bits and pieces of technologies used by MySQL (the product), while MySQL (the company) was purchased by Sun. Then Sun itself was purchased by Oracle, which ended up owning both those previous pieces as well as the core of MySQL. In any case, none, or very little, of the technologies that go into MySQL were developed by or at Oracle. And even if something was, since MySQL, being GPL, comes with full blown, official patent licenses to all employed technologies to anyone who downloads, uses, changes or redistributes it (when one selects to GPL-license something, one also selects to license all patents one owns that go into that something), it'd be free for the use no matter what. So, they're in the clear. Those guys left Oracle to continue working on what they were doing before even joining Oracle in the first place. Nothing, absolutely nothing, requires them to stay with Oracle, or play in Oracle's interests. And Oracle itself is fine with it. Because they explicitly said so in the GPL license. Which they could have changed if they so wished, but didn't.