OSCON, Pt. 3.1: MySQL's Day In The SunOSCON, Pt. 3.1: MySQL's Day In The Sun

On Wednesday I sat down at OSCON with a slew of people from Sun Microsystems to talk about key parts of their empire, both new and old. First up was Zack Urlocker of MySQL (whom I'd observed at the Monday Participate 08 panel), one of the newest additions to the Sun galaxy, and an acquisition that's caused a great deal of worry amongst existing MySQL users.

I asked about what people have been saying about MySQL's acquisition by Sun: "It's generally been very positive. I think Sun has a good rep for open source in recent years, thanks to OpenSolaris, Java, and so on, and I think that's earned them a kind of credibility they didn't have before. The one point that we want to underscore, internally and externally, is that Sun is not just open source but supportive of all the platforms that we run on, including all the different programming languages. Sun may have invented Java, but they're not going to compel anyone, us included, to be Java-only. These were things we cemented with Jonathan Schwartz early on."

Not long after the purchase and the brouhaha about the possibility that some MySQL components would be closed-source, Ingres piped up about the limitations of a bifurcate licensing model.

"We always try to operate in a transparent way -- sometimes that means there are discussions that are uncomfortable in public (like what happened earlier in the year, as you cited). We also try to adhere to OSI-approved licensing.

"Sometimes there are vocal minorities that can never be made happy. We try to stick to our principles with both open source and business, and with over 12 million users it's hard to please everyone. Not long ago, I cut my hair and donated it to Locks of Love, and as soon as I mentioned this I had someone slamming me, saying "They're not a good charity"! [The New York Times had an article claiming that they were not a terribly efficient charity.] Everyone certainly has an opinion. We explain our rationale whenever we can, but we do have a commercial as well as an open source agenda."

Whenever a philosophy normally confined to the margins goes big time, it's hard not to feel like it's become something else.

"When a revolution is successful, it goes far beyond the abilities of the initial revolutionaries and goes mainstream. I don't know that open source is yet mainstream but it's getting there, and we owe a huge debt of gratitude to Linus [Torvalds] and Richard Stallman and the others -- but the issues that the mainstream users of IT care about are sometimes more pragmatic than the revolutionaries. And that's OK. That's a sign of success.

"When companies can make money off something, that's when it becomes sustainable. [A] certain number of people will believe in these things for ideological reasons (and we do have some of that), but that's only one model -- Apache does this -- and it's not always economically sustainable for growth. Apache is a great project, but you need other economic forces for it to be mainstream. Having companies like IBM embrace open source is fantastic. Having Red Hat and Sun support customers using open source is part of the hallmark of success."

I opined that people often hear too much about the open source philosophy from the extremes -- it's either a godsend or a disaster -- even if they're using it in their PDAs or cellphones and don't even know it.

"I think there's truth to the idea that moderate open source advocates need to speak out. We chose the GPL -- we felt it was the best license for now -- but we weren't fanatical about it; there are other licenses out there. If we could help proliferate open source and build a business and we could use the GPL, great. But if there was a different license that worked better, we'd use that (for instance, the GPLv3). We didn't think v3 was all that significant for what we doing, and Linux itself is GPLv2. We were fairly pragmatic about doing what was right for our customers and users.

"At this con there's various hard-core groups that are all about ideals, but this isn't the only open source conference; there are CIOs and CTOs who might go to LinuxWorld and find that more practical. You have to have different venues, licenses, business models, and an open approach to all of them."

On the subject of the proliferation of OSS licenses: "I think there's some truth to that being a problem, but I don't think it's as big a problem as it's made out to be. In the closed-source world, every product is licensed differently, and nobody asks all of them to use the same licenses. There are a handful of highly prevalent licenses (GPL, BSD, Apache), but I think there's fewer active ones. I think over time more companies will abandon vanity licenses and pick one from the stock OSI list that fits. For a while it seemed like every project had its own custom license, but I think we're moving beyond that.

"Sometimes we've had people say 'Why don't we do the Fedora model?' Well, we're at such an early stage of OSI adoption that there needs to be a conscious level of experimentation -- business models, licenses, products -- in order to foster the next generation of innovation. And that thrashing around is OK. We want to encourage that level of experimentation. Maybe it puts a bit more burden on IT to understand the subtleties of these licenses, but the payoff is high and the effort is not exorbitant.

On whether there's a broad difference in feedback from paying vs. free users: "It's such a diverse audience that I don't know if there are broad distinctions between the two audiences. I would definitely say there's distinctions between corporate IT and startup/Web/hobby folks. The most extreme difference is in terms of how regular people want to take updates; the folks who run high-traffic sites will pull source code from our repositories on a daily basis if they think there is a fix that will buy them some increase in scale. In a corporate environment, they may only update every 6 months (except maybe for a security issue that impacts them directly).

"We have noticed a high correlation of Java usage among paying customers (middleware, e-commerce, transactional systems, finance, telcos). We see a lot of Ruby on Rails in community usage, but not in corporate. PHP is spread pretty evenly between camps. The more mission-critical the app, the more likely they will want to make sure they have MySQL Enterprise Monitor; for a casual hobby-type thing, that's not as useful. And sometimes people start off two guys in a garage (Twitter, Facebook, Youtube), and over time it becomes pretty mission-critical with hundreds of servers."

What's cooking now?

"MySQL 5.1 is in its final release candidate stage and should be out next month. A couple of key things that we've added include partitioning. We looked at Oracle and SQL Server as example, and implemented features that both of them use. We also added row-based replication, which makes it easier to scale out horizontally). And there's a 10% to 15% performance gain overall.

"MySQL Workbench (which we announced in the spring) is great for working with inherited databases, where you don't even know what the schema is. We are also adding stuff like Upgrade Assistant (5.0 to 5.1) to the Enterprise Monitor. Other things coming in the fall include load balancing across multiple servers, and Query Analysis -- this is a function which looks at all queries, tracks time spent on each, and generates reports about them. We're looking for ways to make it more intelligent over time. I don't think you can ever replace a DBA, but you can definitely make his job easier."