How Big Blue became an open-source company.
News that IBM
is buying Red Hat is, of course, a significant moment for the
world of free software. It’s further proof, as if any were needed,
that open source has won, and that even the mighty Big Blue must
make its obeisance. Admittedly, the company is not quite the
behemoth it was back in the 20th century, when “nobody
ever got fired for buying IBM“. But it remains a benchmark for
serious, mainstream—and yes, slightly boring—computing. Its
acquisition of Red Hat for the not inconsiderable sum of $34 billion,
therefore, proves that selling free stuff is now regarded as a
completely normal business model, acknowledged by even the most
conservative corporations.
Many interesting analyses have been and will be written about why
IBM bought Red Hat, and what it means for open source, Red Hat,
Ubuntu, cloud computing, IBM, Microsoft and Amazon, amongst other
things. But one aspect of the deal people may have missed is
that in an important sense, IBM actually began buying Red Hat 20
years ago. After all, $34 billion acquisitions do not spring
fully formed out of nowhere. Reaching the point where IBM’s
management agreed it was the right thing to do required a journey.
And, it was a particularly drawn-out and difficult journey, given IBM’s
starting point not just as the embodiment of traditional proprietary
computing, but its very inventor.
Even the longest journey begins with a single step, and for IBM, it
was taken on June 22, 1998. On that day, IBM announced it
would ship the Apache web server with the IBM WebSphere Application
Server, a key component of its WebSphere product family. Moreover,
in an unprecedented move for the company, it would offer “commercial,
enterprise-level support” for that free software.
When I was writing my book Rebel
Code: inside Linux and the open source revolution in 2000, I
had the good fortune to interview the key IBM employees who made
that happen. The events of two years before still were fresh in
their minds, and they explained to me why they decided to push IBM
toward the bold strategy of adopting free software, which ultimately
led to the company buying Red Hat 20 years later.
One of those people was James Barry, who was brought in to look at IBM’s lineup in
the web server sector. He found a mess there; IBM had around 50
products at the time. During his evaluation of IBM’s strategy, he
realized the central nature of the web server to all the other
products. At that time, IBM’s offering was Internet Connection
Server, later re-branded to Domino Go. The problem was that IBM’s
web server held just 0.2% of the market; 90% of web servers came
from Netscape (the first internet company, best known for its
browser), Microsoft and Apache. Negligible market share meant it
was difficult and expensive to find staff who were trained to use
IBM’s solution. That, in its turn, meant it was hard to sell IBM’s
WebSphere product line.
Barry, therefore, realized that IBM needed to adopt one of the
mainstream web servers. IBM talked about buying Netscape. Had
that happened, the history of open source would have been very
different. As part of IBM, Netscape probably would not have released
its browser code as the free software that became Mozilla. No
Mozilla would have meant no Firefox, with all the knock-on effects
that implies. But for various reasons, the idea of buying Netscape
didn’t work out. Since Microsoft was too expensive to acquire,
that left only one possibility: Apache.
For Barry, coming to that realization was easy. The hard part was
convincing the rest of IBM that it was the right thing to do. He tried
twice, unsuccessfully, to get his proposal adopted. Barry succeeded
on the third occasion, in part because he teamed up with someone
else at IBM who had independently come to the conclusion that Apache
was the way forward for the company.
Shan Yen-Ping was working on IBM’s e-business strategy in 1998 and,
like Barry, realized that the web server was key in this space.
Ditching IBM’s own software in favor of open source was likely to
be a traumatic experience for the company’s engineers, who had
invested so much in their own code. Shan’s idea to request his
senior developers to analyze Apache in detail proved key to winning
their support. Shan says that when they started to dig deep into
the code, they were surprised by the elegance of the architecture.
As engineers, they had to admit that the open-source project was
producing high-quality software. To cement that view, Shan asked
Brian Behlendof, one of the creators and leaders of the Apache
project, to come in and talk with IBM’s top web server architects.
They too were impressed by him and his team’s work. With the quality
of Apache established, it was easier to win over IBM’s developers
for the move.
Shortly after the announcement that IBM would be adopting Apache
as its web server, the company took another small but key step
toward embracing open source more widely. It involved the Jikes Java compiler that
had been written by two of IBM’s researchers: Philippe Charles and
Dave Shields. After a binary version of the program for GNU/Linux
was released in July 1998, Shields started receiving requests for
the source code. For IBM to provide access to the underlying code
was unprecedented, but Shields said he would try to persuade his
bosses that it would be a good move for the company.
A Jikes user suggested he should talk to Brian Behlendorf, who put
him in touch with James Barry. IBM’s recent adoption of Apache
paved the way for Shield’s own efforts to release the company’s
code as open source. Shields wrote his proposal in August 1998,
and it was accepted in September. The hardest part was not convincing
management, but drawing up an open-source license. Shields said
this involved “research attorneys, the attorneys at the software
division who dealt with Java, the trademark attorneys, patents
attorneys, contract attorneys”. Everyone involved was aware that
they were writing IBM’s first open-source license, so getting
it right was vital. In fact, the original Jikes license of December
1998 was later generalized into the IBM Public License in June 1999.
It was a key moment, because it made releasing more IBM code as
open source much easier, smoothing the way for the company’s
continuing march into the world of free software.
Barry described IBM as being like “a big elephant: very, very
difficult to move an inch, but if you point the elephant toward the
right direction and get it moving, it’s also very difficult to stop
it.” The final nudge that set IBM moving inexorably toward the
embrace of open source occurred on January, 10, 2000, when the company
announced that it would make all of its server platforms “Linux-friendly”,
including the S/390 mainframe, the AS/400 minicomputer and the
RS/6000 workstation. IBM was supporting GNU/Linux across its entire
hardware range—a massive vote of confidence in freely available
software written by a distributed community of coders.
The man who was appointed at the time as what amounted to a Linux
Tsar for the company, Irving
Wladawsky-Berger, said that there were three main strands to
that historic decision. One was simply that GNU/Linux was a platform
with a significant market share in the UNIX sector. Another was
the early use of GNU/Linux by the supercomputing community—something
that eventually led to every single one of
the world’s top 500 supercomputers running some form of Linux
today.
The third strand of thinking within IBM is perhaps the most
interesting. Wladawsky-Berger pointed out how the rise of TCP/IP
as the de facto standard for networking had made interconnection
easy, and powered the rise of the internet and its astonishing
expansion. People within IBM realized that GNU/Linux could do the
same for application development. As he told me back in 2000:
The whole notion separating application development
from the underlying deployment platform has been a Holy Grail of
the industry because it would all of the sudden unshackle the
application developers from worrying about all that plumbing. I
think with Linux we now have the best opportunity to do that. The
fact that it’s not owned by any one company, and that it’s open
source, is a huge part of what enables us to do that. If the answer
had been, well, IBM has invented a new operating system, let’s get
everybody in the world to adopt it, you can imagine how far that
would go with our competitors.
Far from inventing a “new operating system”, with its purchase of
Red Hat, IBM has now fully embraced the only one matters any more—GNU/Linux. In doing so, it confirms Wladawsky-Berger’s prescient
analysis and completes that fascinating journey the company began
all those years ago.