Interviewee: Marc Miller
In this interview, we spoke with Marc Miller about his views on the current state of open source software. Marc works for Advanced Micro Devices (AMD), and in January, Marc took on a role as the open source software evangelist in the AMD Developer Outreach organization enabling Linux kernel and application developers to develop optimized code using both AMD and 3rd party tools and resources. In his role as a software Alliance Manager for AMD 2001-2006, Mr. Miller played a significant role in developing a Linux marketing strategy with a focus on integration of AMD technology with software tools developed by the open source community and industry partners. Throughout his career at AMD, Marc has been a key contact for open source developers wishing to work with AMD, and has been an open source ambassador for AMD, helping to coordinate outbound and inbound communication between AMD and Linux developers.
In this interview Marc talks about:
- Challenges that open source companies have when competing with proprietary software
- Advantages of open source
- Choice also has disadvantages
- Is open-source anti-capitalism?
- The PR value of being open-source
- Open-source is less idealistic and more pragmatic today
- The future is blending open-source and proprietary software
Richard: First off, I have to ask what you think about the idea of Microsoft funding a blog that looks into the differences in delivery model between open source and closed source. Do you think it’s possible that that blog would be unbiased? I mean, we’re doing our best to make it that way, but I’m wondering if you think we can do it.
Marc: There’s a lot of suspicion right now in the open source community. In the aftermath of the Novell-Microsoft pact, it appears that nobody quite understands what Microsoft is up to and what their game plan is. Certainly the community is extremely skeptical about anything that Novell or Microsoft does at this point based on past experiences, for example with the Samba and Mono communities.
That said, while a blog, focused on open source and sponsored by Microsoft would be treated with skepticism by certain segments of the open source community, there would also be people interested in reading it. Everybody wants to understand Microsoft’s angle on Linux.
Richard: I figured that. I figured there’s going to be healthy skepticism.
Sean: When you speak, Marc, about what the angle is, are you asking what Microsoft’s angle is with Novell?
Marc: Yeah, and the way you look at it will depend on who you are. If you’re a large company like AMD that values both their Microsoft and Novell relationships, we’d be looking at that blog for opportunities where our customers can benefit from the marriage of the two.
One of the key things about the Novell and Microsoft pact was Steve Ballmer getting up on a box and saying that this is to protect Novell’s customers from being sued for use of Microsoft patented IP. That has a lot of people very frightened.
For example, people will be looking in those blogs for what the caution areas are. What products might I get burned on if a court decides Microsoft’s claims on software IP are legally defendable?
Richard: That’s interesting. So, with that in mind, I have a few basic open source questions that I hope you’re willing to address and give us your opinion. One is, what are the challenges an open source company faces when competing with closed source companies?
Take a look at someone like Canonical. All of their software development efforts have been to get Ubuntu into the hands of as many users as possible. They make Ubuntu Linux available at no charge, under open source licensing mechanisms. Their profit model is currently based on support services. For someone like Canonical it’s a question of does the open source community develop products where the user might be willing to pay something for a service like support?
Richard: So, it seems to me that the biggest asset a software company has is the source code. What traditionally has been called trade secrets, intellectual property, or secret sauce.
Marc: In the closed source case, yes.
Richard: It’s an interesting idea to make money by giving your main product away.
Marc: The GPL completely allows for an open source program to be sold for profit. But it also allows multiple copies of that to be made and redistributed without royalties. So, it’s a real interesting area. Since it has been difficult to turn a profit from just selling a GPL application, traditionally most companies that back open source projects use the software itself as an enabler for whatever service or subscription that they plan to sell. In some cases these companies also sell a closed source product that compliments the open source projects.
Richard: So, what do you think are the most compelling advantages to open source over closed source? What is open source able to do much better than closed source can do, in your opinion?
Marc: Open source allows for freedom of innovation in the absence of a driving business case in a lot of cases. Take the Debian community, which has emphatically refused to adopt the same way of doing AMD64 architecture support the way that other Linux distributions have. To give you the ten-second synopsis, other Linux distributions support running 32-bit and 64-bit code in the same environment. Debian felt the way the directory structure was handled in other AMD64-supporting distributions was the wrong approach, and they felt they needed to adopt a structure that would prepare Debian Linux for future innovations. So they elected to support an approach that was different from what other Linux distributions were taking and refused the quick fix implemented for business reasons. I use this example to illustrate that, in the open source community, everybody has the freedom to innovate according to their own interest and ethics, business justification or not. I respect the Debian community for holding to their values. In the closed source space it’s a much more restricted audience; the freedom to do what a particular developer feels is the right thing just isn’t there. People are still confined by ethics, but those can be overridden by corporate objectives and customer requirements.
For example, products sometimes are shipped before they’re truly ready just to stay on schedule. Debian is famous, if not notorious, for delaying a release until the software is really ready and no known critical bugs exist. If a bug is missed prior to release, open source methods allow opportunities for the community to fix the bug after the fact.
In open source software, anyone can disagree with a corporate direction and then redistribute their own version which has things written the way that they think it ought to be. This promotes a freedom to innovate that goes beyond what you can often find in the closed source model.
Richard: How does open source keep from having endless slightly different distros? Is it Darwinian? The things that really work and get picked up live on, and those that don’t just sort of peter out?
Marc: The biggest strength and the biggest weakness of open source is the freedom of choice. Michael Dell has been quoted in the press prior to the May 1st announcement that they were going to start shipping Ubuntu. Each time he’s quoted about responding to when are you going to start shipping Linux, when are you going start supporting the community, he says, "I’m fine with supporting the community. Which one? Which one do I pick?"
Every Linux distribution has a different philosophy. Look at Red Hat; they have had a very long history of preserving backward compatibility and ensuring stability of the products such that when their enterprise customer upgrades, they’re not going to have to recompile their kernel and reinstall a bunch of apps.
Look at the SUSE case; they try to take a much more balanced approach to include innovation in cases where it might be a minor inconvenience, but they believe that the customer will appreciate the value that they get out of the product for that minor inconvenience.
Now look at Debian; I once suggested that Debian try to encourage enterprise ISVs like Oracle to integrate their software with Debian Linux as a way to expand their visibility and market share. A Debian developer responded, "Who cares about market share? What we care about is open source." Everyone on that discussion list thread agreed there’s no reason to make Debian more compatible with Oracle databases because Oracle isn’t open source. Debian’s mission is an open source revolution across all software segments.
So each Linux distribution, besides having a different code base, has a different philosophy behind it. And that creates a lot of choices for the user and that’s a big strength, but that choice complicates the decision too. There are two main product lines in the Windows release – a desktop/workstation line, and a server line of products, and one can even argue that those two product lines are more similar than dissimilar as they leverage many of the same operating system components. There are 359 open source operating system distributions according to Distrowatch, most of which are Linux.
Richard: It seems to me in doing research on this subject, I run into an awful lot of open source advocates who seem to be open source advocates more because of the anti-capitalism sort of flavor to it, at least in some corners of open source.
That seems to be a common voice, although it doesn’t seem to be a common motivation of all of the open source projects I’ve looked at. It looks like there’s an awful lot of people writing open source trying to figure out a way to make money on it. They believe in open source for maybe collaborative reasons, maybe for development cost reasons or whatever, but they want to make a profit.
But this anti-capitalist message seems to be sort of pervasive in one from or another through almost everything I’ve read. Do you think that that hurts the open source community in the eyes of people who are agnostic in terms of development paradigm? Do you think some people are turned off to open source because they see it as sort of anti-business, and they aren’t?
Marc: The attitudes of the open source community have changed over the years. If you look at the Linux community of the late ’90s, you’ll find that’s very much the case, to the point when we launched x86-64.org I asked that all of the corporate logos be removed from the website. Since then we’ve brought them back because the open source developers realized that without a corporate motivation and corporate sponsors that there’s a lot that they’re prevented from doing. Many open source applications are hosted and mirrored on either boxes in a corporation’s data center or using corporate sponsorship money. Linux Symposium wouldn’t happen without corporate sponsorship. Linux-supporting companies look after the interests of the open source communities in standards bodies and industry forums that developers don’t usually have access to.
Also, a lot of corporate sponsors pay developers to do what they would do anyway. And those developers have become grateful to those corporations in a way that has lessened the anti-capitalism trend.
Sean: Let me ask you a follow-up to that, and I know exactly why Richard called it an anti-capitalism bent, but do you see that as the most appropriate phrasing for it? I mean that’s kind of the quick draw phrase that everybody uses, but is it anti-capitalism or is it anti something else?
Marc: I’d say it started off as anti-capitalism because the feeling was that most corporations A) produced strictly closed source software and B) didn’t understand the open source model well enough to be any good to what people like Richard Stallman were trying to promote. So, it was an anti-capitalism setup.
But as I say, modern-day, most of these well known developers are sponsored by a business. So, a lot of that trend has died down.
Sean: So what do you think it’s kind of evolved into? Is there still kind of an anti- bent to the whole community? I mean is it kind of anti-control, or anti-throttling of innovation, or what would you stick there?
Marc: The Linux community is a lot more pragmatic about software IP than they were 15 years ago but long term, open source enthusiasts still are driving toward elimination of software as a proprietary product. It’s a freedom of knowledge movement, and much in the same way that you can go to Wikipedia and read about any topic that someone has contributed knowledge to. In the software case, that freedom of information is creating software instead of encyclopedia articles, and the content creates a new market. You can still make money off software creations, without making money from the software itself. Perhaps that’s services for the software, perhaps it’s a subscription to access premium content.
Open source enthusiasts feel that they are resisting controls over innovation and strive to obtain freedom to innovate in a way that’s beneficial to the customer, regardless of business case. It’s not anti-business but does challenge many existing business models.
Richard: In other soft product areas, like music, publishing, and stuff like that, you have a war between the copyright holders and those who think that stuff should be free. Do you think that closed source companies have a right to protect their source, or do you think somehow they’re subverting the greater good?
Marc: My opinion is that they have every right to do with their products whatever they need to do. If the software author also supports an open source movement, that’s a great thing but the open source model is not for everyone. In Eric Raymond’s "Cathedral and the Bazaar," there’s a whole section about reasons to open source, and that if your product does not meet all those criteria, then maybe you are better off keeping it closed. Often a solution stack is incomplete without certain pieces of software that happen to be proprietary.
For example, people buy a solution stack that includes Oracle and SAP because they derive value from the closed source applications, not because it’s built on top of an open source infrastructure. The open source operating systems (such as Linux) offer great features too, but today, most OS licensing and services revenue is driven by the application workload inherent in the solution stack the OS was bundled with, not necessarily the OS or benefits of open source.
Richard: So, you don’t think open source is in opposition to closed source, but rather just a different paradigm?
Richard: OK. That’s cool; that’s interesting to know. I was trying to run down project-development leads on OpenOffice.org, and I think a little bit over half of them are employed by Sun. I’m curious, speaking non-altruistically, what does a company gain by sponsoring open source?
Marc: Besides the points I’ve already covered, they gain favoritism as an open-source-friendly company, that developers will even develop for their closed products because they believe they are contributing to the greater-good effort.
Richard: So, it’s sort of developer PR at some level? I’m sure not all of it, but there’s some, "We’re good guys, look, we’re sponsoring."
Marc: From a business perspective, that’s right – that’s what many of them see. The ability to leverage innovations from the community is also attractive, but the contributions aren’t guaranteed in the same way that the developer PR is.
Richard: That’s interesting.
Sean: For about seven years I’ve been involved in deep, technical efforts around evangelism, and I’ve certainly seen closed source companies realize and economic gain from sponsoring and supporting a community.
For a company that’s got a mainly closed source model, and they embrace open source model on the side, do you have any kind of anecdotes or examples of places where the company looked back and saw gain?
Marc: Probably the closest thing that I can think of is Adobe’s decision to allow projects like OpenOffice.org to include PDF generating tools, and their decision to make PDF a completely open specification that they encourage other people to put into their products. This has benefited Adobe’s PDF software products, even though Adobe’s products are closed. So, by creating an open standard, they’ve enlarged the market for themselves, and by supporting open source efforts, they’ve gotten a lot of mileage in the market that they sell into.
Richard: Last week Adobe open-sourced Flex, which was a free product. What, do you suppose, they gain by open sourcing it? Do they gain lower development costs, because they’re going to get free collaboration?
Marc: Well, I’m not very familiar with Flex, but suppose that they did the same thing with Acrobat Reader, which is another free product that they give away. They have already developed the market to the point that PDF is ubiquitous. If you want to make a document that is guaranteed to be readable on just about any platform out there, whether it’s x86, Apple, or a smart phone, you generate it in PDF.
If they decided to make an open source Acrobat Reader, that allows a lot of projects to tap into it and borrow code from it; it promotes the whole integration story of plugins and addons that go beyond the original plugin interface for the application. By giving their developer community the freedom to innovate, the development direction would be managed by the community.
You asked about lower development costs. The cost only goes down measurably when the community contributes something that the “copyleft” holder was going to pay for. Open source software lets a lot more innovators in the door to help with software development, but sometimes additional resources have to be found to do the work that no one contributes for free. The point that the open source community is trying to make is that open source innovators are shared among many open source projects. It would be expensive to recruit each of those developers under contract, and even more expensive to maintain that software over a long period of time.
Indeed, many projects have benefited from the open source model. Open source is not a guarantee of success any more than having great technology. There has to be demand for it, and from the right audience or no one will show up to do the work.
Richard: I see. A couple weeks ago, I ran into the CSI company that was started by Stuart Cohen. It’s sort of open source. The paradigm is that CSI hires developers, and managers as part of a geographically centered development teams. Companies who want a particular piece of software pay for CSI to manage that effort.
The idea is you might have half a dozen companies who want to have some software package. They pay CSI to develop it, and they also may make some of the developers on their staff available to the management of CSI to contribute to that project.
However, the thing that interested me the most is that was that once the software package is developed, it may not be re-licensed as open source. Rather, it’s going to be the decision of the contributing companies how to re-license, or even whether to re-license it. The presumption, I guess, is that they’ll go ahead and let it be available through some sort of open source model, where it can be modified and redistributed without cost, but they don’t have to.
It looks like a blending of open source and closed source. You’ve got this shared cost and collaboration of open source, but you don’t necessarily have the other main underpinning of open source, which is that the source code is available to everybody.
Do you think that that’s where we’re going, into different sorts of paradigms that blend proprietary and open source methodologies, and do you think that’s good or bad?
Marc: Though they would prefer everything to be open source, open source developers are beginning to accept that there are certain things that are difficult for a company to expose without getting them into potential legal trouble. They don’t like it, but they’re understanding. Overall, I think it’s a good thing.
Richard: So, it sounds like open source is drifting away from highly idealistic to more pragmatic; still holding onto their ideals, but being more pragmatic about attainability in certain situations.
Marc: There are some exceptions, but for the bulk of the open source community, yes, they’re becoming a lot more pragmatic about things. They still prefer to hold onto certain ideals, and continue to direct everything in that direction, but most developers are fairly pragmatic about it.
Richard: So, I had a couple of interviews with some guys at Microsoft focused on ensuring security for Microsoft products: Michael Howard and James Whittaker. Both of them, separately, said that the biggest advantage that closed source has, in almost every area—architecture, implementation, quality assurance, and documentation—is the ability to mandate a process. Microsoft has the ability to say, "This is part of the job, and you have to do it," whether it’s unit testing, adhering to coding standards, security reviews, or whatever. The implication was that would be hard to put together with open source because all you have is the code, you aren’t certain of the steps the developer went through to produce that code.
Is there good adherence to process in open source projects? Is it something that’s improving? Can you comment on that area, in general?
Marc: The open source community has code control procedures and models of their own. For example, in the LinuxBIOS/OpenBIOS project, you can submit a patch to the mailing list only once you’ve gotten one of the key maintainers to review and approve the code. So, there are still very much the rules for developing that code and processes to be followed. Some of these methods probably emerged from source code control models originally used in closed source settings, but in the spirit of open source, why change or reinvent what’s already working well?
The rules for submission seem to vary a lot for each project. You can have quite a range of different quality control mechanisms. But, usually, there is a quality control mechanism there.
Richard: Yes, there has to be. You can’t release software unless you have some kind of quality control. It seems like you have to have some sort of gatekeeper over the code base, or it’s going to deteriorate in a hurry.
Marc: Without someone moderating the process, you end up with wars of someone changing a bit of code, and another person changing it back. Endless back and forth. Code control also guides the direction of development so that community goals can be met.
Marc: A lot more of, as you put it, the blending areas particularly as Linux becomes more widely used in different server and desktop environments. For example, in multi media areas we’re seeing a lot of companies that create set top boxes that use a combination of closed source and open source software in embedded devices. It’s an area where businesses are realizing there’s money to be made with the combination of open source software and closed source IP.
Virtualization is a big area of expansion. A lot of companies want to make money off of virtualization, sometimes for security, and sometimes it’s to increase the range of compatibility across operating system environments. There are both closed and open source options, such as KVM, VMWare, and Xen. Each innovates using different philosophies. So, this is an area that I’m watching with interest.
Device drivers are also an area of much attention. The Linux kernel community has been very emphatic that they will not accept any closed source code into the Linux kernel. And yet the way that you develop drivers for Linux is to include it in the kernel. For example, the FCC has given 802.11 wireless networking devices a certain range of frequencies that may be used. Many of those devices are essentially programmable radios. The hardware vendors don’t want to open source that because that allows someone to then misuse the product and gain access to frequencies that the FCC wouldn’t approve of. Sometimes in order to reveal specifications, licensed IP used in the hardware design has to be revealed.
The drivers I’ve just described can never be part of the Linux kernel, and yet Linux users want to be able to use all these devices. In recent months, just really January of this year, certain key people in the Linux community are starting to realize that this prevents a lot of hardware companies from participating in the way that they would like to, and that though hardware vendors have nothing against open source models, for one reason or another they simply cannot release their code. I’ve been watching a number of efforts to make it easier for hardware developers to produce Linux drivers. I hope this topic is discussed during the next USENIX Kernel Summit and that the maintainers find a way to prevent Linux from being stifled further on the basis of compatibility problems with new technologies.
Richard: Do you see Linux making inroads into the desktop market? Do you think that they can gain a significant part of that?
Marc: Client computing models are changing from traditional desktop and laptops, and with these changes there are new opportunities for Linux. Different types of desktop applications or use models also offer an opportunity for Linux.
I think in time we will see Linux emerge in the same way – having strengths in certain types of client devices. I don’t know what that strength is going to be yet for Linux but many people in the community are exploring what it should be. Today, most ISVs only want to deal with one desktop operating system, and that’s Windows.
Today there are a number of ways, including virtualization, that you can run Windows apps in a Linux environment. As that improves, I think we’ll see a lot of ISVs soften to that idea. And, as they look at some of the things that we’re developing for Linux, they may decide that it’s better to develop a Linux app that will also run under Windows than it is the other way around.
As the Novell-Microsoft Interoperability Lab continues to grow their efforts I can definitely envision Microsoft having to run Linux apps inside their environments much the same way that we have those today in Linux.
Richard: It seems to me that products like OpenOffice.org and other open source products that are becoming strong, mature applications in the traditional desktop applications space have to work to mitigate Microsoft’s grip on that market. In the past, when Windows first started to take over the desktop, it was because that’s where all the applications were. You could put Linux or something else on there, OS2 or whatever, but all the mature applications were Microsoft applications, and they had to run on Windows. And I think that’s sort of how they got that grip on the desktop market. It seems like the first step toward making that a more competitive market is through the applications phase.
Marc: Yeah, I agree with that.
Scott: It looks like the TCO battle has been fought to a stalemate. Microsoft has Microsoft sponsored reports saying they have lower TCO. On the Linux side, there are reports sponsored by Linux supporters saying that Linux has lower TCO. Do the communities find any of these reports credible? Have you come across any truly independent TCO reports that you’ve thought were especially good? Is it even reasonable to expect that the finding of some TCO report will be borne out in the specifics of your organization?
Marc: In the independent studies I’ve read, per machine, the costs are pretty much the same for a supported machine if you compare Windows to a truly equivalent Linux implementation, but this fails to take into account the workload processing power costs. In the open source model, I have visibility into what code is slowing down my e.g. mail server, exclude the things I don’t need to have running, and make the machine more efficient. Without that optimization, more machines would be required for the same workload regardless of the operating system in use, which adds to the cost.
Richard: Marc. Thanks for taking the time to chat with us.
Marc: Thank you.