OSCON, Pt. 4.2: openSUSE's Eleventh Hour (And Twelfth, And Thirteenth...)OSCON, Pt. 4.2: openSUSE's Eleventh Hour (And Twelfth, And Thirteenth...)

Aside from having one of the niftier names in the industry, Joe "Zonker" Brockmeier has a pretty nifty job, too: He's the openSUSE Community Manager at Novell, where he oversees the folks that help make what will ultimately turn into the next version of SUSE Linux Enterprise. I grabbed a few minutes of his time to follow up on things I'd talked to him about back at the Red Hat Summit.

"The uptake for openSUSE 11 has been pretty good. Obviously, being open source, we don't have perfect numbers to determine who's updating. When we first released 11, though, we saw huge 20 MBps traffic through Akamai (which we had to use for the first couple of days), versus something like 12 MBps for 10.3. And obviously we saw a lot of positive reviews, too. Generally feedback has been very good -- a few people have discovered that KDE 4.0 is still rough, but we were up-front about that fact and included KDE 3.5 as a fallback."

That led into some discussions about KDE's mistake in labeling 4.0 as a 4.0 release and not as a beta or a release candidate.

"I think the KDE project was very candid about the capabilities of KDE 4.0, but the naming was a bit of a boo-boo. Quite frankly -- yes, I don't think people read the rationale; they saw the 4.0 and assumed from there everything was solid."

What's behind openSUSE's strong KDE following?

"Probably because SUSE has had a strong KDE contingent historically. Maybe it's been because we're also popular in Europe, but we know that SUSE users tend go about 70% KDE. That may be changing with this release; I would be curious to see how many people are using 4.0 vs. 3.5.

"That's one of the few things where open source doesn't have an advantage; we don't get the marketing numbers that proprietary software vendors do. If we had that kind of data, we'd know more about our demographics."

Getting those kinds of stats without making people feel like their privacy is being invaded is tough, though.

"One of the few ways to get accurate pictures is through updates. We're starting to try and get some data from that to see how many unique IPs are hitting update servers, so we're not tracking specific users."

I drew an analogy for how the record industry saw its sales charts change overnight when they went to a point-of-sale model for units sold, rather than just counting units shipped to distributors and retailers. Clearly, there needs to be more endpoint tracking of usage.

"Right now, one of the projects that we're working on that we've included but is not the default is Fedora's package for assessing such things. We are working on making it the default, and I can't guarantee that it will be, but that would solve two problems -- one being measuring these things and the other being having two distros doing it instead of one. Also we'd be looking at not just installs, but whole hardware profiles. That way we can go to vendors and make a case for them shipping open drivers for a given piece of hardware."

What's next on the slate for SUSE?

"We don't have a complete picture, but obviously there will be more improvements in yast (we made huge strides in 10.3 and 11). They had a workshop for yast and came up with a bunch of new things to work on. For example, using yast as a Web service and maybe configuring things that way. We'll have newer versions of KDE and GNOME, of course, and as is the way we do things, this next release of openSUSE (11.1) will also be the release for SUSE Linux Enterprise 11."

What distinguishes the community from the enterprise version?

"A lot of the people who are paid to work on Enterprise also work on Open, and we don't leave it entirely up to the project to come up wit the feature set. The Enterprise guys have ideas about what should be in there, so we have certain features that are pretty much dictated from that side. As Open becomes more independent, we have more leeway to do what we want. Since we released the first 1.0 of the openSUSE build service it now becomes possible for external developers to work on equal footing with internal developers -- submit changes and fixes, and so on."

Perhaps a better question would be, what constitutes an enterprise-level feature in SUSE? (I pointed out that the more different folks I talk to who have community and for-pay versions of a product, each one has a different criterion for what they should charge for.)

"In our case, the enterprise version is not distinguished by functionality, but by ISV and OEM certifications, hardware certifications, and most importantly the length and life cycle of support. We assume people running Open have a tolerance for updating every two years, vs. the enterprise, which is supported for much longer. For us, we wouldn't take any time to strip any features out of, say, Apache for openSUSE.

"This model has worked well. There isn't a lot of overlap. Really small businesses or home users, or developers doing leading edge (not bleeding edge!) work will run Open. People who need long-term support and enterprise needs will get Enterprise.

"To that end, we target a couple of demographics -- Linux enthusiasts and developers, and new Linux users. One of the goals of SUSE is to be the easiest Linux to use and obtain. One reason we have a retail box with software that isn't free (like Flash, etc.) is because people who are switching away from Windows will want that functionality legally, and not want to jump through hoops to get it. It's not hard to get MP3 functionality in Linux -- for someone who has experience! And it's not that hard for someone who's determined and still doesn't know. But a lot of users get annoyed when the have to jump through hoops in the first place.

"I think that like it or not, you understand that the reason a certain distro can't just ship MP3 support are legal reasons. But the average person that downloads and installs Linux just sees that as a problem, and they don't understand why that limitation exists at all. It's just a defect to them.

"There's an interesting psychology I've noticed: When you tell someone Linux is free, they think, 'Well, it might not be worth much.' Tell someone 'Here's an OS with all this and it's only $60, and you get 90 days of install support,' they think, 'That's a bargain.' People pay for convenience.

"There's also the threshold of what's reasonable in terms of trading money for software. Is it reasonable to pay $60 for an OS with a full range of apps, as opposed to Windows OEM versions plus Office, which together can easily cost as much as the hardware itself? That's unreasonable to me.

"I've owned a Mac running OS X and there are a lot neat little apps that are well designed, but you very quickly get the sense of being nickeled and dimed to death. A text editor: $15. App that does something else: $20. To get the same functionality out of a Linux distro, you've had something like $200 or more chiseled out of you.

"If we want to se Linux on the desktop, we need to meet users halfway. We can't keep a purist attitude exclusively. The vast majority of computer users are not motivated by software ethics."

Or by software, period, I commented.

"The other thing I would suggest is, looking at people who make a case to switching to Mac and then going back to Linux (because Mac is not free). The real lesson is not that there's thing wrong with free software -- but that people have to see it firsthand. You cannot convince them with mere arguments."

We digressed for a bit about the odd practice of version numbering, and its fundamental arbitrariness. "In our case, with each major version update, the .1 iteration of that becomes the next major version for Enterprise. We'll release 11.1, 11.2, keep working on stuff, and 12.0 will be the start of Enterprise 12. More often, the versioning is arbitrary as any other criteria. The shift between 10.3 and 11.0 was fairly dramatic because we made major changes to the installer and packager. But we might make major changes between 11.2 and 11.3, but because our numbering is based on what we do with Enterprise, then we're not going to bump it up to 12. Each vendor seems to have its own numbering.

"I think we should stop using version numbers and just name everything after the developer's cats. OpenSUSE Fluffy, that's the next version!"

No crazier than Hardy Heron, if you ask me.