Interview: OSDL chief Stuart Cohen - Part 1

How are things going for you at OSDL?

Very well. I've been there about 18 months. By almost any market share measure, Linux gets double-digit growth quarter-on-quarter and has tremendous success. That part's good.

When I got here we had 23 member companies. We've about 60 now, and will probably finish the year with 65-70 and reach 100 next year. Lots of people are interested in participating in what we're doing. Our membership plans are $10,000, $100,000, $500,000 and $1m. Our budgets [and] reserves are going up.

We were [also] primarily technical. But as Linux moves more mainstream, we're probably equally technical, marketing, business and legal issues. We're also broader, [covering] Linux and both open source and proprietary Linux-based applications that run on top, because that's the reality of the software portfolio out there.

We'd like to see more people participate, more workgroups, and we'd like to see more Europeans involved [as] we're mostly US and Asia. We've offices in Tokyo and, just opened, in Beijing. We've a single UK-based employee in Europe, but we're [planning to] open a new office in Europe in the next few months.

Where's that likely to be?

Three candidates: Brussels, because of the EU, the UK and Germany, which is doing the most with Linux. We've got a little more work before we figure that out.

We've a European customer advisory council of about 20 participating Global 2000 companies. We've similar councils in the US and Japan [and] we'll have one in South America starting around [January 2005].

What are the OSDL's main successes to date?

Technical: Obviously with [Linux creator] Linus Torvalds and [kernel maintainer] Andrew Morton we're doing our fair share of code development, and the subsystem maintenance and performance testing work we do is significant.

Business: some thought leadership we've been doing has been very helpful.

Legal: the white papers, the legal defence fund, education, ideas around a prior art repository and work we're doing around trademark, patents, licensing and copyright is all very important [although] we haven't come out very much on that [yet]. Then there's our work in the telecoms market [with carrier grade Linux].

What are joining companies really looking for?

In most cases they're focused on data centre, carrier grade or desktop [Linux] and [identifying] the technical and market inhibitors to knock out for acceleration to take place. In the US and Europe people want to participate in those workgroups.

Some Asian companies join because the open source development community is much smaller for us to bridge between their country or organisation and the rest of the world's development communities.

I'm sure there are some who think that if they join the OSDL they're going to talk to Linus and Andrew all the time, which doesn't really happen, and some that just do it for networking with other companies and the inexpensive market research.

Are these inhibitors being tackled?

The major ones are. There's a lot of kernel things that [version] 2.6 solved. Performance on 8/16/32-way is dramatically different. Some technical things are dramatically better in 2.6 than 2.4.

There are more issues around the ecosystem - ISV porting, application certification, binary compatibility amongst distributions - the programs and the libraries that make up a new, broader definition of Linux.

What are your priorities now?

Probably number one is ISV [independent software vendor] software adoption and the ability to run software across multiple distributions with a single image.

It's very important to make sure that the value proposition is strong for users and ISVs, and that Linux doesn't go the way of Unix where there's a forking of the operating system at distribution level.

I heard of a user with Red Hat and Novell SuSE Linux, because one critical application was only certified for Red Hat with other key software only available on SuSE. Is that an example?

That's not as big a problem to me as let's say Oracle is the database behind those applications. If Oracle had to support different versions on SuSE and Red Hat, so the customer had to support and pay for two versions, the value proposition starts to deteriorate. That's a bigger problem.

[From your example] my guess is that the application provider said the level of support the user wanted had to come from a single distribution. If it were Oracle, we're finding they don't care which distribution because they'll hold [Oracle] accountable for the results.

Is getting end users and resellers on board a priority?

It's important that they participate. We've got 60 to 70 companies in Europe, 90 per cent of which are vendors, so I'd like to get a little more in balance with users. But we have three councils with people participating [from] 55 different companies. They provide the business feedback we need to balance the vendors' viewpoint. When we got those up and running I felt a lot better about the balance.

What we're about is data. The money's nice, but not critical. We need their input, their involvement. We're getting a lot of value from [these councils].

We have data centre, carrier grade and desktop workgroups, and special interest groups around networks, security, storage and the like. We use the councils to provide feedback and validation on things we're thinking about in those workgroups. So we present our latest results, thinking and maybe specifications and prioritisation. We put their feedback into our working groups, which helps us define projects.

Are there any specific new projects?

One is what we call 'the OSDL Working Set'. It's on our website and readily available. [It] looks at the programs and libraries commonly used by big companies [that] go with the kernel to make up a suite that application companies need to test, certify, port or validate. Linux is not just the kernel any more. It's bigger. So that working set is an important piece for us.

Under the GPL licence, changes feed back. But there's always a time lag, isn't there?

Absolutely. And there are things that the Linux kernel project doesn't care about. But it's a program/library/package used by broad number of users that an ISV has to be compatible [and] compliant with.

Would these need to be certified for a series of named distributions?

It would depend. At minimum we want the definition on which programs and libraries we need. Then there needs to be some testing [and] certification. Some code probably needs to be written against the LSB [Linux Standard Base] standard we're working with today.

If that [LSB] definition is broad enough to include these programs and libraries, and doesn't have to come from us, great. We just care that it's out there giving a strong value proposition: the ISVs can follow it, and everybody's using it.

So these ISV functions have to be available on the different distributions?

Correct. Binary-compatible. We're [also] doing work around binary regression testing with a lot of industry members. That's another project on our website you can go and see. It's pretty interesting [and] also new, so nobody's paid much attention to it. It's a set of code written off the LSB standard that provides a layer for ISVs to port to above the kernel itself. It's safe to say that could be a considerable project in terms of porting.

You want the operating system to be the same between the hardware and application layer, and not leave any room for the mistakes that were made with Unix where different vendors had different Unix configurations, and all of a sudden locked everybody into their own hardware or applications.

What are Linux's biggest immediate challenges?

On the desktop there are very attractive solutions [for] fixed function office workers. Managing your own email, calendar, browser and applications works very well with Linux today.

[But] a lot of areas around mobile computing and mobile professional group calendaring, replication and synchronisation need work.

Server Linux is moving along well. Obviously Linux is under attack because [of] its success. The beauty is that this creates competition. But as you have winners and losers, sometimes you get irrational behaviour.

So will it just be a straight slug-out between Microsoft and Linux in the end?

Yes. I don't know how much of a slug-out. If you look at the open source software out there today, [it's] going to gain ground against proprietary software but it's not broad enough to cover a lot of the proprietary portfolio.

I tell people, if you have 100 employees who've been around more than two years you probably have some proprietary code you couldn't get rid of even if you wanted to, because it's the fabric of your business. Some of that's Oracle, SAP [or] PeopleSoft. It's been around forever, and will be. And you probably have some legacy code you wrote yourself in-house you're going to need. So there's a lot that won't go away.

What do you think will happen over the next five years? Desktops and mobile for instance?

Novell, Red Hat, a lot of companies in China [and] a number in South America [are creating Linux desktops]. In the US they're looking at desktop applications running on Linux. I think the desktop - I really mean a thin client, fixed function office worker - will grow pretty rapidly.

[But] there's a lot of work to be done [for] mobile professional group executive and administration [staff] who share email and calendaring. If you want dominant growth you've got to get it on the mobile side. I think 45 per cent of PCs being sold in the world today are laptops.

Are you planning a mobile initiative?

No. Right now we're looking at usage models within the desktop. We're looking at mobile [but] it's probably last because it's the most work.

I clearly understand what you're saying that you can't have mobile lag too far behind, otherwise you'll never get the desktops [but] it's a low percentage market share today. You have to have short-term success to get long-term opportunities.

With mobiles, you have an odd mix: Windows dominant on PDAs, Symbian coming up from phones, Java sitting on anything, some Linux models and some mobile standards.

It's highly competitive. Windows is the clear winner on the desktop looking back. The question is: can any of those disruptive technologies significantly go after [the mobile] market?

Will business drive down to the consumer for the desktop?

What people do in their businesses will drive what they do at home. The good news for Linux is that the web piece, email piece, calendaring piece - if you manage your own - is pretty good. You only get yourself in trouble when you and your secretary both share email or calendaring, or you and a number of different people on different servers try to look at a number of different calendars. [They're] probably the weakest things, and as a consumer you won't use any of those.

But shared facilities are vital for businesses, so fixing this should be a high priority.

Absolutely. You can do it [now] but it's got to be integrated with your BlackBerry [or] Treo, cellphone and everything else - not just your laptop. There's replication and synchronisation work to get done. But I think business will drive consumer behaviour.

Have you a time scale for the desktop becoming mainstream?

A couple of years. Two or three years. You can go a long way on the desktop without getting to the mobile professional, [maybe] to 20 per cent of the market without a single laptop user. So it's a significant upside on the desktop before you run into the problems, and before you [do] a lot of them will get fixed.

What's the big message you would like to see coming out of OSDL?

It depends on the audience I'm talking to. In general I'd say that Linux has gone mainstream on the server and there's a lot of work going on around the world to make that true on the desktop as well.