Original URL: http://www.theregister.co.uk/2010/02/20/open_source_decade/

Open source - the once and future dream

Ballmer, trolls, and the next ten years

By Gavin Clarke

Posted in Software, 20th February 2010 00:01 GMT

For some, Oracle's $5.6bn purchase of Sun Microsystems was good news for open source. After all, a mega tech vendor has acquired a hugely popular open-source database product and project: MySQL.

But at the same time, as the European Union's anti-trust probe wore on, costing Oracle's chief executive $100m a month, and MySQL co-founder Monty Widenius ran a vocal web and press campaign against the deal, fears began to grow that Oracle and other big software companies would set the on-going drama as a reason to refrain from getting involved in open source. Some worried that the advances of the last ten years would come tumbling down.

Larry Ellison munches dolphin

As it announced its strategy for its Oracle and Sun products last month, the database giant said that - far from being deterred by the experience - it will remain actively involved in open source. "If we someday acquire another company and that company happens to have open source products, we would be perfectly happy to pursue that opportunity," chief corporate architect Edward Screven told press at Oracle's Redwood Shores, California campus.

Oracle's swallowing of MySQL closed a decade that saw free and open source (FOSS) leave the geek or enthusiast hinterland of IT and join the business mainstream, occupied by the likes of the database giant itself. Fittingly, it was Oracle in 1998 that helped usher in that acceptance through its pioneering decision to put its database and application server on Linux in 1998. The move was political, a way of sidestepping Microsoft's control of server operating systems through Windows.

Since Oracle's endorsement of Linux, the kernel and operating system have penetrated everything from high-end data centers to handheld consumer electronics. Linux runs stock exchanges from New York to Tokyo. It powers Sony's Playstation, Tivo recorders, and Samsung TVs. It's on Google mobile phones, tens of thousands of Facebook servers and 90 per cent of the virtual images in Amazon's EC2 cloud.

Late last year, ABI Research estimated that nearly one-third of the 35 million netbooks sold in 2009 would ship with Linux. Meanwhile, 11,000 lines of code are added to the kernel each day, an indication of the kernel's increasing utility in daily life and also one the sheer number of participants working on the project.

Linux pulled open-source applications along with it. FOSS moved into middleware and applications: JBoss, SugarCRM, JasperSoft, Talend, and Terracotta all emerged in markets that had been dominated by closed-source products with expensive licenses from big vendors that included IBM, Oracle, SAP, Informatica, Microsoft, BEA Systems, and others.

Open-source code produced Drupal, MediaWiki, Joomla, WordPress, Facebook, Firefox, and WebKit. Open source helped kill BEA, Borland WebGain, and - yes - Sun, which were either too slow in responding or failed to respond at all to the idea of free code and open systems. These companies could not find a way to justify charging people for their tools or middleware when people were gobbling up good-enough of better-than-expected systems from the open-source community.

Another landmark of the open-source decade was that mid-way through the 2000s - and after many years of much talk - Sun released most of Java to the community. This was important because Java, once the upstart challenger in the world of programming, had by the mid 2000s become a mainstay of enterprise systems and mobile devices. People actually relied on Java. The only sticking point was - and remains - the Java test compatibility kits (TCKs) that prove compatibility and that Sun did not release to open source.

So where does the chain reaction go from here and where will FOSS be in the next 10 years? The challenges are significant, and largely they are issues unresolved from the previous 10 years: intellectual property, companies' struggle to make money on a decent scale, tensions between the idealists and the more commercially minded members of the community, and the struggle to convince corporate users to participate and actually give back.

They are challenges that, if solved, could see FOSS and Linux hit a whole new level in the next 10 years, but if unresolved permit uncertainty to fester in a way that ultimately limits their business and technological growth.

Ballmer fallout

Patents and intellectual property are FOSS's great unsolved unknowns. The industry might shake its head and tut each time Microsoft's bullish chief executive Steve Ballmer lobs a grenade with his talk of potential violations of his company's patents in Linux, but this is just theater. The real action is to be found away from the headlines, on a day-to-day basis, where corporate lawyers and opportunists fishing for easy money lodge claims of violation against those using open-source.

Such suits are not specifically aimed at Linux per se. They're part-and-parcel of doing business in a litigious country under a system of patents in software. Linux and FOSS simply get caught up in the game.

Often, deals are reached out of court to avoid the expense of fighting a long and costly battle in the weeds: It can cost up to $5m to defend the average patent litigation case in the US. Little wonder that such cases are settled out of court, either for payment or in return for licensing or payment of revenue from the product concerned.

During 2009, it was Microsoft versus Tom Tom over an alleged patent in FAT, and the case was settled privately without the sides disclosing the terms. Before that, Microsoft extended patent protection coverage to Novell and Linspire under separate agreements. Significantly, Red Hat's always held out against Microsoft's coverage.

Open source provides rich ground for patent trolls and litigants. Code might might find its way into, say, a Cisco Systems router, because some third-party down the line didn't understand - or didn't bother to check - the terms of the code's license. Patents, meanwhile, are a staple of the industry: IBM is consistently the industry's single biggest patent go-getter. But, worryingly for open-source, Microsoft is catching up rapidly, as part of a policy of licensing and then monetizing patents.

The problem is that each new agreement effectively fences off an area of FOSS behind a legal arrangement on a company-by-company basis, potentially limiting the use of FOSS. The open-source community is well aware of this and companies such as Red Hat have founded organizations like the Open Invention Network to buy up patents and defuse the lawyers.

Unless the system changes, or until litigants and trolls go away, the prospect is for more of the same in the next 10 years. This prospect could see FOSS slowly hemmed in and its creation and pass-on-rights slowly strangled as claim that someone's IP is contained in the code are quietly settled in a way that people outside the case cannot verify.

The incredible shrinking patent portfolio

One hope against such a future is that those prosecuting cases can't make enough money from their claims, making the pursuit of claims uneconomical. Linux Foundation executive director Jim Zemlin pointed to the fact those buying patents portfolios have found they didn't earn as much money as they'd hoped and are pruning portfolios.

Zemlin believes the inability to monetize patents has given companies a new taste for collaboration instead of registration and prosecution.

There's also the prospect for a change in the way US patents are awarded. Zemlin thinks the US Patent and Trademark Office (USPTO) will become more "rational" and institute reforms specifically targeted at software patents now that it's headed by David Kappos. Kappos is a former vice president and assistant general counsel for intellectual property at IBM who managed IBM's patent and trademark portfolios, protecting and licensing intellectual property worldwide.

"I do think that from an administrative side the person appointed to the head of the USPTO has a good deal of experience around software. So it's to that degree that the USPTO can become better trained at discovering prior art and on the issue tougher software patents," Zemlin told The Reg.

When you consider that a former IBMer is helming the USPTO, combined with the fact the world's largest software company - Microsoft - is aggressively trying to copy IBM by registering as many patents as it can, you realize that the status quo may be here to stay. Microsoft's FAT patent, for example, remains intact despite numerous community attempts to challenge it on the issue of prior art. FAT is implemented on billions of machines, from PCs to cameras, meaning rich pickings for lawyers working on volume if not size of individual settlements.

Ultimately, Zemlin believes companies that aggressively enforce patents will lose support of developers and that any software they have will lose out in the marketplace. That's dangerous because without developers, their platforms won't get applications, and without applications, they won't attract the customers they need to make money.

"I see honest consternation from developers who avoid creating products on platforms produced by people who go around beating everybody up with their patents," Zemlin said.

The other way to cash in

Patents are not the only way people have tried to make money from Linux and FOSS during the last 10 years. The story of the 2000s was of people charging for services to support Linux and FOSS.

While projects and products saw growing numbers of downloads, the practice didn't necessarily translate into sales, and for some companies, this was a case of trying to find creative ways to close the stable door after the horse had bolted.

Linux companies - Red Hat and Novell in particular - fared well. In terms of downloads and cash, Red Hat came to dominate the last decade: Red Hat had 62 per cent of the business Linux market compared to SuSE, according to IDC.

And, because Linux was driving business applications, businesses were willing to pay the companies for support and services. Linux closed the 2000s on 12 per cent of the server operating system market based on revenue in the third - most recent - quarter of 2009 measured by IDC.

The only thing that separated Red Hat and SuSE were their individual performance. Red Hat performed relatively consistently with annual and quarterly growth, but while SuSE increased its market, Novell - its owner - was beset by losses and restructuring.

The picture was less clear cut on applications and middleware. JBoss application server and the MySQL database were two of the decade's biggest successes. This came by way of downloads and disruption rather than in terms of making money for the companies supporting them.

BEA gets JBossed

JBoss helped first to frustrate and then to undermine BEA's thousands-of-dollars per CPU price model by giving developers a free application server for development and rollout. It was partly responsible for dispatching BEA into the arms of Oracle when its management couldn't respond. Meanwhile, MySQL was a runaway success among web developers, used in conjunction with Linux, Apache, PHP, Perl, and Python (LAMP).

Red Hat spent $350m buying JBoss, but four years later, JBoss is making an estimated $100m a year. Meanwhile, in 2007 MySQL began throwing the kitchen sink at users in an attempt to make any money with the all-you-can-eat Enterprise Unlimited license priced $40,000 that came with 100 hours of free consultative" support plus monthly updates and service packs.

By November 2008, nine months after Sun closed its $1bn purchase of the database company, MySQL was growing - at 48 per cent a quarter that was faster than Sun's existing systems business - but it was still only pulling in $37m.

With a new decade here, the question is what's next for the new and existing crop of FOSS services companies like Alfresco, JasperSoft, Penthahlo, SugarCRM, and Lucid Imagination. Do they grow and expand as independent companies and start to challenge the top-tier IT vendors in their respective markets? Will they be bought? Will they finally call it a day due to lack of growth or the sheer hard work of trying to turn downloads into gold? In the case of the latter, it could be their venture capitalist backers who finally tire and move on.

State of FOSS

Today's FOSS companies claim some big business customers and growing revenues. They are mostly believed to be making around $15m to $20m annually, while Alfresco is believed to be puling in $40m. These companies are private and do not break out their numbers.

Typically, in the software industry, the goal is to break the $1bn mark and begin another growth after that. That's certainly the goal of Red Hat's chief executive Jim Whitehurst.

Jaspersoft CEO Brian Gentile believes it's only a matter of time before companies like his hit an inflection point when revenue from subscribers cover expenses and at least become cash-flow positive - a figure he puts at between 800 and 1,000 subscribers. Gentile predicts Jaspersoft and others will hit that inflection point during the next two years.

"There are a number of use about to hit that phase," Gentile said. "We just have to exercise our patience."

Rivermuse co founder and open-source veteran Dave Rosenberg believes that while open-source companies can grow, it's more realistic to see them make no more than $100m in annual revenue and feels that the magical $1bn mark is a stretch goal. The reason? The nature of open source - the fact that code is already out there and you must persuade customers to pay you to support something that their own techies are comfortable with and capable of doing.

"This is the blessing and curse of open source - it downsizes the market in many ways," Rosenberg said. "As you make this software available for free, they are sucking the life out of the market. A market that might have been worth $200m will becomes $100m, because customers and are getting more value from it."

Rosenberg believes success when it comes has been where open source companies have challenged a clear player on price - like BEA, Oracle or Documentum - and offered users a very similar experience, so they've not had to adapt from what they already know. This has allowed users to switch very easily at fprecisely the time when it came to renewing their very expensive enterprise license with their existing IT provider.

Diversify, diversify, diversify

Another issue inhibiting the ability of opensource companies to hit the $1bn mark is that many have a limited number of products - often one core product or service - that they are only slowly building out. Open source theorists might dismiss this as charmingly old school and place their faith in the ability to make money from the long tail of demand and from volume markets. But the IT giants they are challenging and emulating are armed with massive arsenals of product that helps them rake in billions of dollars each quarter or each year.

Of course, the really big money earners in the world of open source are those who already have other products or services and are not centered around open-source. In the case of Oracle, it was the database. When it came to IBM's 2005 purchase of open-source application server Gluecode, it was the rest of IBMs vast applications and services portfolio. Others, like Cisco or Hewlett Packard, are using open-source code in their networking equipment or running on their PCs and servers.

Jean-Noel de Galzain, chairman of the program committee for the Open World Forum 2009, was blunt in his assessment. "Today a few open-source companies are really emerging in terms of pure revenues...there is a lack of economic credibility around the open-source business model," De Galazin told us. "Does open-source really bring and open-source business model or does it influence the existing business model?"

He believes this will be the big dilemma for open source in the next decade: Can it function inside purely open-source companies or does it work best inside companies that make their money through other - potentially closed-source - means?

When whales eat open source

Big Software will continue to buy open-source companies, but the last decade offers little guidance for future acquistion trends. Some of the deals of the last decade were head scratchers. Citrix bought XenSource for $500 million, Yahoo! acquired Zimbra for $350m, and VMware acquired SpringSource for $420m, in addition to Sun's purchase of MySQL and Red Hat's acquisition of JBoss.

At least MySQL fit well with the Sun strategy and its CEO's beliefs in software, while JBoss dovetailed with Red Hat's aspirations as a platforms company. But Yahoo! already had mail before it owned Zimbra; smf the logic behind VMware's purchase of SpringSource remains open to debate.

Financially, the numbers are all over the map and are supported by acts of faith, not the slide rule. The best, strongest justification has been to acquire a community that the new owner can, somehow, monetize over time. It's the same kind of thinking that saw companies in the first decade of this century believe they could charge for services from supporting code already in the community.

"The numbers are out of wack," Rosenberg said, pointing in particular to VMware's purchase of SpringSource. "But compare that to what VMware would have had to have done - they could never have been able to get the community of Spring users."

What would be the fate of open-source companies bought by closed-source giants?

One risk - hardly unique to IT - is that the acquired company is stifled by the corporate culture of the buyer or - worse - deliberately killed to stop it from taking money away from the new owners' existing products. That's been a concern of many MySQL users over Oracle's purchase of Sun: Oracle has gone out of its way to avoid accusations of killing MySQL by saying it'll be kept as a separate sales and development entity.

Jaspersoft's chief executive doesn't think it's a bad thing for open source outfits like MySQL to be bought by closed giants like Oracle, but hes worried in case their open-source methods and models get "snapped out. Even the largest software vendors like Oracle have to figure out how the open-source methodology plays into their business models," Gentile said.

"I like to think Oracle's acquisition of Sun and MySQL will be an accelerator of Oracle being a good citizen in the open-source world," he added.

But regardless of whether a company is bought or whether there's value in its products, there's the question of open-source licensing.

License for madness

There are 66 "official" open-source licenses recognized by the Open Source Initiative (OSI). Each recognised license is slightly different in terms of specific geographic jurisdictions or restrictions while some are specific to companies, products, or institutions.

Of the licenses, GPLv2 is the most popular for open source, but arguably, it's not the most "business friendly" - meaning companies can't alter code or keep their changes or make money off of them. That leaves many working with the Apache Software Foundation's license, which is considered "friendly."

The system is confusing and acts as a disincentive to adoption. It is potentially open to abuse as people ignore the license and make the changes they want, or put open-source in the products they want without respecting the actual terms of the license. Furthermore, de Galzain believes that different licenses could hurt the international expansion for some of the FOSS companies that want to break that $1bn barrier.

This could prevent a new generation of open-source users in emerging nations and at home as they are deterred from using open source, because of the confusing and complex nature of the different licenses, or because they get burned by experience - either from lawyers or community activists pursuing those who are in accidental violation a license.

De Galzain believes the system of licenses should be made clearer and that the OSI, GPL author the Free Software Foundation, and the Linux Foundation are the ones who should be working on this. "We will have to make this license system clearer for everybody," De Galzain said.

He believes newcomers to the community should work with existing licenses - not create their own, as, for example, Microsoft did. "For a software publisher the best way to have an efficient open-source strategy is to use the common standard... so using GPL or Apache or Affero [for cloud] and not creating a new license," he said.

"For our businesses, if we want to create world wide companies and businesses... we have to use mainly GPL or the Apache license, or BSD-like basic licenses."

The Linux Foundation's Zemlin agreed license proliferation is bad, noting people already have enough difficulty in simply maintaining technical compatibility. He noted, though, the market does seem to be correcting itself around a few licenses - GPL and Apache in particular.

"For the hold outs that insist on creating new licenses I believe there should be education as to why that is a bad idea, but these seem to be rare and the market seems to be correcting itself in this regard," Zemlin said.

The open source lifeblood

Companies come and licenses go, but the thing that really makes FOSS are participants - those who build the code and make changes. Contributors are the lifeblood of Linux FOSS, and the 2000s saw Linux and FOSS attract legions of new contributors as businesses recognized the potential of the software and the development methodology. New contributors meant new features.

The challenge of the next 10 years is in continuing to not just attract more contributors - but to attract one very specific type of contributor: the end-user corporation outside the existing gene pool of IT that dominate.

Many of the contributions to FOSS come from those already in the tech industry. That's either because they have a natural inclination or interest in working on coding projects or have of a sense of self-interest with individuals or companies working on code to projects and products that help them down the line.

But end-user corporations outside the tech sector developed a reputation during the last decade for using open-source and not returning their changes to the community. In some cases, the terms and conditions of employee's contracts mean what ever work they do during their nine to five on the company clock belongs to the company and cannot - from a legal perspective - be simply donated to a community or given away. In other cases, employers would not release changes for fear of betraying competitive advantage.

Red Hat's chief executive Jim Whitehurst - an ex-Delta Airlines man - in 2008 put the subject of needing to get more users from outside the tech community on the map, lambasting the level of participation in open source. He slammed a lack of user-based projects, saying hundreds of billions of dollars are wasted each year as enterprises build their own software.

Bring out your contributors

Michael Tiemann, Red Hat's vice president of open-source affairs and president of the OSI, told us fresh contributors are needed to help bring improvements to code on a small level that can make open-source as a whole better. Tiemann believes there should be greater synchronization between different projects if open-source is to really make a change now that it's hit the mainstream.

"Open source dramatically expands the number of ways people can participate in this incremental improvement cycle, and when open-source is doing its synchronization job these smaller efforts come back to be a major one," Tiemann said.

He also believes those corporations that use open source and open-source methodologies will benefit from this process, because changes in code become synchronized and because software no longer becomes a big-bang release that can go wrong during rollout.

Backing Whitehurst, Tiemann claimed in a whiter paper here (PDF) that $1 trillion is wasted each year in failed IT projects as software runs late, goes over budget, or fails to deliver on the features promised. Last year, the Standish Group reported a quarter of all IT projects are considered a failure because they are canceled or are never used. Forty four per cent are late, run over budget, or are delivered without the required features. Thirty two per cent are considered successful.

Open-source helps, Tiemann says, because those changes on a small scale mean smaller upgrades and less "big" IT rollouts once there's been a major application upgrade.

Tiemann says it's in the end users' own self interest to join in because participation improves FOSS and makes the experience of using FOSS better, because users making the changes they want in the software. "Nothing succeeds like success. We have seen a growing number of companies trying, failing, trying again, failing a little more, and ultimately, the success makes it all worth it," Tiemann said.

"There is an extraordinary potential to be unlashed when people discover how to properly interact with open source, which is not consume, consume, consume, or buy, buy, buy - but by this behavioral ability to be able to inhale, process, exhale - to make participation rewarding."

So participation is good, but how do you convince those consuming to give back or circumvent the corporate legalese that prevents donations to FOSS during the next 10 years?

The cultural shift

Hacking is the easy part. The hard part is the cultural shift organizations must make so their employees can begin participating. That's a slow process that requires positive feedback on the experiences of open source inside organizations and flagging up success stories.

Tiemann notes that IT departments are the ones that are ones least likely to be trusted inside an organization when it comes to advocating greater participation. That's because they are blamed for the project failures of the past and because IT has earned a reputation for being a cost center, not a center of savings. So while IT can advocate, it will likely take the backing of the business, management, or some other force to bring change.

This is a long process, but the Linux Foundation's Zemlin is positive that the rate of contributions to Linux and FOSS in general - and from outside IT in particular - will increase and cease to be an issue during this decade.

Zemlin says that people who don't contribute today will eventually, pointing to how the Linux kernel has grown during the last 18 years. "The benefit of open source is you can collectively maintain open source...I think this is a temporal issue not a systematic issue."

Meanwhile, Tiemann believes the fact so many large organizations are already using open source means we are now at a tipping point - where users outside of IT become contributors. He cited the government of Brazil, where one official told him recently it's running open source "everywhere" but that it can now see "limits" in where it can go by just being consumers because open-source is not giving Brazil all the features it needs. "That's why they need to become co producers," Tiemann said.

"It's as if you got the field ploughed and prepared and maybe even seeded and now it's a question of getting people tending the crop," Tiemann said. "Ten years from now, we will have a huge harvest - more people participating and more features."

The past catches up

If the 2000s were remarkable for being the decade when Linux and FOSS crossed into the mainstream. The next ten years will show whether the last decade was a one-act play or the first chapter in a longer story.

The challenges that were the hallmarks of the last 10 years have not gone away. The issue now is whether Linux and FOSS can solve them or brush them aside and proceed to - as Tiemann believes - touch the rest of society in the way the PC and Microsoft's Windows revolutionized society by putting low-priced computing power in the hands of the masses.

Zemlin takes the example of Windows a step further. Looking to the end of the 2010's, Zemlin draws on the famous vision of Microsoft co-founder Bill Gates for a computer on every desktop and every home that drove that last revolution.

"My vision," Zemlin said, "is to have a computer in every gas pump, X-ray system, cell phone, GPS system, set top box, picture frame, car, logistics system, airplane, DVR, server, super computer and desktop all running Linux." ®