Five Questions With Michael Widenius

volume4 writes "With two MySQL execs leaving Sun in the last week, the internet is buzzing about what is going on at Sun, what is the future of MySQL and what lies ahead for Michael Widenius. Over at Open Source Release Feed, Widenius spoke candidly regarding his split from Sun, the future of MySQL, Monty Program AB, and the open source ecosystem in general."

Comments

Politics as usual

[Anonymous Coward]

people in MySQL management made my life hell; They didn't let me participate in MySQL development, didn't give me resources in doing Maria development and did a lot of backstabbing to make my life difficult.

Do you find the politics in the real world more difficult than the ones in the Open Source community?

Re:Politics as usual

[von_rick]

I think this thread is about 5 question for MW that he has already answered :)

Back to what you said: I think it was mostly the management and less of politics. The role of management is usually to dictate terms that they have absolutely no clue about. A marketing exec managing a SQL functionalities is gonna make the team go sour given enough time.

Re:Politics as usual

[zappepcs]

Whether the speculation on reasons for the two leaving is right or not, there is one thing that can be said: MySQL had enough going for it that Sun paid a tidy sum to get it. Whatever they did to cause the 'break up' it was not a good thing.

Car Analogy: If you buy a racing team; expensive cars, mechanics, drivers... and two star drivers walk out... well, lets just say winner's circle is probably not in your short term future, and the cost of business operations just doubled.

Re:Politics as usual

[sunking2]

Not if you have Ricky Bobby in the pits.

Re:Politics as usual

[mb1]

...or the beautiful death machine that is ShakeNBakeSQL.

Re:Politics as usual

[laa]

I think this thread is about 5 question for MW that he has already answered :)

Fascist! :)

MySQL to be thrown onto the community?

[dov_0]

While it wouldn't be the first of Sun's projects to go open, maybe they will limit their input into the project and give preference to open source contributions. Maybe take things up again post recession.

Re:MySQL to be thrown onto the community?

[ShieldW0lf]

What a funny interview. He talks all about trust and open source, and yet one of the major reasons I have always avoided open source is because I personally distrusted the two major developers.

I remember reading countless times how this or that missing feature wasn't needed, and how it was bad practice to use it in the first place. Then, next version, they'd brag that they had it. They would ignore referential integrity, but hide that fact, and call their bugs "Gotchas" or "Features". They would claim that having intelligence in the database and reduce traffic across the wire was bad practice, and that you should move entire result sets into the middle tier and filter it there in their forums, and on and on and on.

If they had been forthright about what compromises they made, what the strengths and weaknesses of their design were, and been prepared to acknowledge that there were tasks it wasn't fit for, I might have put it to more use. But at the end of the day, I couldn't trust the developers not to engage in misleading behavior, so I stayed far away from it, and used PostgreSQL instead.

In the end, the developers tried to pull a DivX Networks/Project Mayo type of move and rip off the community, and this was only reversed when Sun bought them out.

It's good to see that those lying, thieving bastards are no longer involved with the project. Particularly since I am now obligated to use their bastard child at work.

Re:MySQL to be thrown onto the community?

[RichardJenkins]

If the future of MySQL lay in community developer contributions - why bother contributing to MySQL directly? It's all (mostly) still available under the GPL. Form it, and create a community version.

Sun would never import code back into MySQL unless they owned copyright to it though.

Re:MySQL to be thrown onto the community?

[ShieldW0lf]

I agree -- any database that does not support foreign keys (referential integrity) is really not a database -- it's just a toy. Use Oracle, DB2, or Postgres for any 'real' work.

See, that's the thing. There are narrow niches where that's a strength, where a little bit of garbage data in there isn't going to cause any harm, and the speed mitigates the weaknesses. And if they had been upfront and forthright about the capacities of the project, that would have been fine.

But when you come across another "not-a-bug-a-gotcha-or-a-feature" every other month that you didn't know about because they were forthright, and you're forced to work with projects that were incredibly poorly designed because people trusted the developers advice and did everything wrong... there isn't any trust there. You couldn't feel safe using it for anything significant unless you're actually getting right in there and hacking it yourself and seeing what it does, because you know from experience that they'd flat out lie to you if they thought it would increase their market share.

Good he could sacrifice a good 30 seconds

[ergo98]

My question to him would be "Why? Why send so many naive and misled followers back to Microsoft Access level technology and choices when we should have taken those lessons and moved forward?" MySQL, like PHP, is one of those mistake technologies that thrived despite itself, and when you go to the root of it you find someone saying "I knew nothing about the technology, but just started building from scratch, re-making the mistakes every other product made 20 years earlier".

Re:Good he could sacrifice a good 30 seconds

[kilodelta]

I'm not sure I agree with you at all. LAMP has its place. Put it this way, I've administered MS-SQL, Oracle and MySQL databases, I'll take MySQL any day.

Re:Good he could sacrifice a good 30 seconds

[ergo98]

I'm not sure I agree with you at all. LAMP has its place.

I said nothing about Linux or Apache. LA certainly has its place, though if you look back in the early years of Linux -- remember back when you had to go through a bunch of kernel make files and hand edit all of your driver settings and so on -- there was the same sort of attempt to make lemonade out of lemons that we've seen with MySQL's multi-year mistake. At the time you had to actually compile every driver into the kernel, and this was heralded as a model of efficiency and custom suited kernels, and so on. Then it gained the loadable driver ability, and that farce was quickly discarded. We've seen the same thing with MySQL, where as it finally gains functionality that competitors have had for decades, we're finally hearing the end of the ridiculous "But it's *good* that it has laughable integrity" arguments.

Put it this way, I've administered MS-SQL, Oracle and MySQL databases, I'll take MySQL any day.

Well since you put it that way...

Re:Good he could sacrifice a good 30 seconds

[musicmaker]

Two words: Explain Plan
Three More: Share nothing cluster
A date: '00-00-0000 00:00'
Two More: Silent truncation
One acronym: MVCC
Result: Nobody in their right mind uses MySQL.

LAMP has it's place, it's at the bottom of a trash heap. Ever tried to write business objects in PHP? What about dependency injection? Database abstraction? (let's face it PDO is a joke). Hell even prepared statements are a pain in PHP/MySQL (only exist in mysqli, and the implementation is horrible). AOP? You can't even do connection pooling for goodness sake because they turned it off in mysqli, and you need your head read if you are using the regular mysql libraries where the solution to injection attacks is to escape quotes and pray. Do you know how long it takes PHP to parse 80,000 lines of libraries every time a script runs because there is no persistence between requests, so PHP has to parse everything over for each request.

MySQL where foreign keys are silently ignored if you forgot to set your table engine to InnoDB. Where aggregates don't work right, where self referencing updates don't work, so you have to write a program to do what other RDBMSes can do in a single statement. Where your table names are case sensitive, but your text matches aren't.
Where you don't have sequences to generate globally unique ids, where bit fields work like a boolean half the time and char half the time. Where mysqldump locks half your database and doesn't get everything by default which you find out too late because you didn't know any better.

Apache where the recommended default for MaxClients is 256, which anybody with a clue knows is insane for dynamic websites, but most sysadmins put in anyway. PHP that hasn't been bothered to update itself to work with a threaded Apache that has been around for a decade.

I could go on for ages and ages on this stuff. I mean there are SO many issues with LAMP, it's a minefield. LAMP fails when you need it most, when traffic starts getting heavy.

OR

you could use a system that separates components into libraries and interfaces, allows you to modularize, allows database independence, makes testing easy, has static typing so the compiler can catch 80% of problems before they ever get executed. Has AOP, has IOC that isn't insane and is used by more enterprise shops that anything else.

Re:Good he could sacrifice a good 30 seconds

[anomalous cohort]

Killing MySql won't send people to MS-Access. It is PostGreSql (which was Sun's original favorite OSS RDMS) that stands to benefit from the death of MySql. I had predicted this a year ago.

Re:Good he could sacrifice a good 30 seconds

[DragonWriter]

Put it this way, I've administered MS-SQL, Oracle and MySQL databases, I'll take MySQL any day.

Perhaps, but what about PostgreSQL?

Re:Good he could sacrifice a good 30 seconds

[kilodelta]

Never dealt with PostgreSQL. Did dabble a bit with DB2 though.

Re:Good he could sacrifice a good 30 seconds

[sarkeizen]

There's lots of comment that could be made about that diatribe too.

First off I'd say that some of this criticism is "Well it doesn't have *buzzword*" mixed with a few statements you are likely not in a position to argue along with a number of things you got wrong enough to betray some ignorance of the subject you are criticizing and at least one instance of "Well the default config doesn't suit what I'm doing". The fact that you have to reach all the way down to that makes me question the "I could go on for ages" bit.

Considering that you appear to have completely stupid amounts of emotion invested in your particular choice of tools I won't really bother arguing them all.

One that you mention twice is database abstraction. Personally I'd call DA a double-edged sword. Sure PDO doesn't achieve the level of abstraction that lots of frameworks do (but one might argue that it's not a framework so...) but at least if I hire someone who writes PHP I know they can write a simple join in SQL. I've met huge numbers of people - from professional developers to university students - so mentally locked into a platform that they couldn't do this.

As an aside one thing I will say about the developers I tend to hire is that they have to show proficiency in writing code in a few different languages and perhaps some aptitude in writing code in a language they've never seen before. One thing I find this cuts down on is the amount of time they spend complaining about language X lacking feature-they-love Y which tends to get in the way of doing actual development.

Re:Good he could sacrifice a good 30 seconds

[drinkypoo]

I could go on for ages and ages on this stuff.

You could, but it would be a big fat waste of time, because thousands of people would still use LAMPs for the simple fact that the software costs nothing so hosting is readily available, and there are tons of working content management systems available for free to run on top of them, and by the way numerous commercial sites are doing brisk business with them. So what if they don't fit the needs of HP or Xerox?

Most of your complaints are about defaults. That's dumb. I mean, the defaults are dumb, but you're still a whiner.

Re:Good he could sacrifice a good 30 seconds

[sarkeizen]

We have exactly one production Postgres DB in our environment and we will migrate it to MySQL soon. There's a few things that are drivers to this:

i) DB support on Linux apps - if it runs on Linux it tends to support MySQL but not necessarily postgres (one app I recall running into with this difficulty was wordpress). Because of this the more Linux apps you run, the more you are likely to run a MySQL db which means that even if you standardize on Postgres you will probably break that standard frequently.

ii) Clustering: The landscape may have changed today but at the time we were trying to get a DB cluster running for failover and performance and although MySQL was ram based it was still better than just about every Postgres solution we looked at.

iii) MySQL isn't as bad as people think: Most of the time I hear complaints about MySQL they are things like "It isn't ACID compliant". Which is like 'Welcome to four years ago buddy!'. I'm not about to argue that MySQL is equivalent in every respect to Oracle but that for a wide variety of enterprise applications MySQL is sufficient. So much so that a lot of shops might do well do standardize on MySQL and drag Oracle out for that application that can't be moved to anything else.

Mind you I would like to migrate our MSSQL apps to MySQL as well but there we have the added driver of cost. Not only is the server expensive but clustering seems to, without exception require special more expensive versions of SqlServer and Windows Server. Because of that we are currently using a block-level replication solution for failover and during one of our fire-drill tests the DB was corrupted.
   

Re:Good he could sacrifice a good 30 seconds

[Tuntematon]

Ever tried to write business objects in PHP? What about dependency injection? Database abstraction? (let's face it PDO is a joke). Hell even prepared statements are a pain in PHP/MySQL (only exist in mysqli, and the implementation is horrible). AOP?

I don't even know what the these terms mean so it must mean I can still use MySQL and it has a place after all?

Re:Good he could sacrifice a good 30 seconds

[toby]

Why ask Monty? Why not ask every major web player on the planet why they choose 'Microsoft Access level technology'.

You might find yourself laughed out of the room (as Microsoft would be, if they tried to sell Access to Google, Yahoo, Flickr, Slashdot, imdb, SABRE, YouTube, Wikipedia, NASA, US Census, ... and most major MySQL users).

http://mysql.com/why-mysql/case-studies/
http://mysql.com/customers/

Re:Good he could sacrifice a good 30 seconds

[Slashdot Parent]

you could use a system that separates components into libraries and interfaces, allows you to modularize, allows database independence, makes testing easy, has static typing so the compiler can catch 80% of problems before they ever get executed. Has AOP, has IOC that isn't insane and is used by more enterprise shops that anything else.

I'm curious on this one. When is the last time you swapped out one database for another?

Also, can you name a few good uses for AOP other than trace-level logging and security? I'm not trolling, I just really wish I knew the answer and I've been asking everyone who sounded like they could give one to me.

Re:Good he could sacrifice a good 30 seconds

[ergo98]

Why ask Monty? Why not ask every major web player on the planet why they choose 'Microsoft Access level technology'.

You realize that MySQL wasn't invented yesterday, right? That it has plied its course for some 14 years now.

For about the first 7 of those, it was absolutely horrific, chosen only by idiots who had absolutely no idea what they're doing. Since then, sure it has started to gain features that every other RDBMS vendor has had long before, with each step the naive userbase suddenly coming into realization of what they were missing, finally abandoning the hilarious rhetoric they'd been spouting online since.

Though it's funny that now we're doing it all again with products like CouchDB. I laughed when I watched the video of the Damien guy where he professes to having started with no idea about databases, but why let them stop him? The history of MySQL follows the same sad tale. In both cases, people afraid of real databases just clutch onto their talking points and rhetoric and tell us that their super ridiculous chosen technology is somehow superior in its inferiority.

Re:Good he could sacrifice a good 30 seconds

[Simetrical]

Where mysqldump locks half your database

FWIW, if you use it with InnoDB (and almost any serious MySQL shop uses InnoDB exclusively), mysqldump --single-transaction is lockless.

Re:Good he could sacrifice a good 30 seconds

[crucini]

...and almost any serious MySQL shop uses InnoDB exclusively...

I work at a pretty serious MySQL user, and we use both MyIsam and InnoDB. Properly tuned and used, Isam is faster. Innodb allegedly has the edge in PK-lookups, but my measurements disagree.

Innodb is good if you want Oracle-ish features. Mostly we don't. In fact, the current is flowing the other way; towards things even leaner than MySQL.

We mysqldump the slave, which eliminates that issue. (You can also stop the slave and cp the db files; I think this does not work with Inno). Table locks remain a serious performance issue in Isam; it does take extra effort to prevent lock contention.

Re:Good he could sacrifice a good 30 seconds

[Simetrical]

I work at a pretty serious MySQL user, and we use both MyIsam and InnoDB. Properly tuned and used, Isam is faster. Innodb allegedly has the edge in PK-lookups, but my measurements disagree.

InnoDB also doesn't have to be repaired or checked. MyISAM tables can crash on occasion if you look at them funny, and then you have to spend an inordinate amount of time repairing the tables (possibly days for large data sets). InnoDB supports data clustering, which is extremely useful for disk-bound data sets. And table-level locking in MyISAM is prohibitive for most serious workloads, unless either there are no writes (other than end-of-table inserts) or all of your queries are fast.

I'm curious: how large are the tables at your shop, and how many reads/writes per second do you have? I know that Wikipedia, for instance, would melt if you tried using MyISAM anywhere, and it's no more than a mid-sized MySQL user as far as its database size goes.

We mysqldump the slave, which eliminates that issue. (You can also stop the slave and cp the db files; I think this does not work with Inno).

It does, in my experience (with the usual version-mismatch caveats), but only if you copy the entire MySQL data directory. Some essential parts of the InnoDB data files are shared across all tables and can't be split out by simple filesystem commands.

Re:English is Author's Second Language?

[Precision]

Am I being picky today or is English the author's second language?

Yes it is, iirc he's from Finland.

Five Questions...

[RayMarron]

...three of which were answered "See my blog". Article rated -1 uninformative.

Re:Five Questions...

[quickOnTheUptake]

I got the distinct feeling he was kinda fed up with the interviewer.
I liked the "I just answered that." It takes real skill to get someone to answer like that in a short non-controversial interview.

Re:Five Questions...

[Dynedain]

I disagree. His "I just answered that" response was referring the interviewer/reader to the previous question, to which his answer was basically "Read my blog".

The blog link is a post to a mostly-empty, non-informative wiki. If you aren't going to answer the questions, don't take the interview. This was just a cop-out of "I'm to lazy to explain things to you, figure it out yourself" which is a horrible stance to take when you're trying to launch a new venture and gain interest.

Re:Five Questions...

[musicmaker]

It's yet another demonstration of the professionalism of the MySQL people. Maybe Sun can introduce some level of seriousness about software to MySQL. Oh wait - those are the people that gave us Java; nevermind.

before open source?

[sl0ppy]

David Axmark (the second founder of MySQL) and I released it 'open source like' (this was before open source) in 1995

i think that several people would disagree with that.

mySQL was seen at the time as an answer to mSQL, which was non-free, even to the point of sharing a very similar API.

funny how times and history change.

Re:English is Author's Second Language?

[grouchyDude]

Not a good enough excuse for a pointless article with little information. Hardly seems to justify promotion to the front page.

Re:English is Author's Second Language?

[Anonymous Coward]

And IIRC his first language is swedish.

What's the next big DB?

[EmperorOfCanada]

Obviously not an emergency, but what is going to be the next big DB? Obviously SUN is going to screw up MySQL sooner or later (if not a bit already). I really don't want to develop for the next few years without knowing what DB I am going to end up switching to. Will PG reemerge from its slumber? Will someone fork or branch MySQL? Will SUN go under and MySQL break free (1 billion $ later)? Firebird looks vaguely interesting. So my question would be: If your boss was allergic to MySQL what would your next choice be? My next question is: How will SUN screw up MySQL? Java based trigger language, Build up of annoying bugs, Proprietary high cost "Enterprise" version with critical features, A "rebuild" that completely ruins the whole thing, A tacky marketing program that makes people embarrassed to say MySQL, A new name that costs 3 million or more. Pointless features that support some other SUN marketing effort, ... and my favorite ... Neglect.

Re:What's the next big DB?

[Frankenshteen]

I think the big fork is already available from Google. And with their investment in MySQL, it can't go away anytime soon. http://code.google.com/p/google-mysql-tools/downloads/list

Slashdotted

[rrohbeck]

Coral cache:
http://www.opensourcereleasefeed.com.nyud.net/interview/show/five-questions-with-michael-widenius-founder-and-original-developer-of-mysql

No URL in your comment...

[drinkypoo]

...even though it's about "See my blog." Your comment is -1 uninformative.

Re:No URL in your comment...

[RayMarron]

The URL is present in the FA. I'm simply normalizing the comment data.

Finland?

[Dr. Cody]

I though he was from Christmas Island.

Oh, wait...

How well do you know MySQL?

[toby]

InnoDB does support foreign key constraints. (And I'll mention that it has ACID transactions as well, since you may not know that either.)

Re:How well do you know MySQL?

[ShieldW0lf]

InnoDB does support foreign key constraints. (And I'll mention that it has ACID transactions as well, since you may not know that either.)

Well, actually, InnoDB was made by another group and is currently "owned" by Oracle if I'm not mistaken. Regardless, I remember when they didn't have these things, and they constantly put forth that they were some backwards way of doing things that was unjustified and that if you were doing things right, you didn't need them. This isn't about features. It's about trust.