Author: Joe Barr
In 1995, Behlendorf was working at Wired magazine, which had one of the first non-academic Web sites. In those days, everyone used the NCSA (National Center for Supercomputing Applications) Web server, and most users were concerned about its future. NCSA engineers began being hired away by a new company called Netscape in the summer of 1994, and the early-movers on the Internet worried that with both a server and a browser, Netscape might “own” the Internet.
Brian explained that users had been fixing bugs and sending the fixes back to NCSA all along. As response time from NCSA grew longer and longer, individual users of the NCSA server began swapping patches between themselves. That patch-swapping activity on mailing lists later became the Apache Software Foundation.
Eventually, it began to look like NCSA was never going to release a newer version than 1.3, and the task of keeping up with all the individual fixes to that, and overlaying them, and keeping track of who had what, became a major pain point. Some of the users — including Behlendorf — decided it would be a good idea to issue a new release themselves.
They debated about whether they should issue a 1.3.1 release, or a 1.0 release under a new name. They looked at the license with the code, and Behlendorf paraphrased it as saying, “Do whatever you want with this, just make sure that you give NCSA credit before you give it to somebody else.”
The only thing left was the question of what to call it. Contrary to popular belief, Behlendorf said he chose that name because “I wanted a name with a connotation of ‘take no prisoners’ and suggested it on the mailing list. Another soul on the list immediately pointed out that since they all had patches for the server, Apache was also a very good pun — if there are any such — on ‘a patchy server.'”
Eventually the Apache project began doing a lot more than putting in bug fixes to the old server. It redesigned the core so that new features could be added without major trauma. If someone suggested a new feature, they were told, “Hey, that’s cool. Go away and write it to meet this API and if we like it when you’re done, we’ll add it.”
According to the Netcraft Web Server Survey, Apache has become the dominant server on the Internet today, with more than twice as many active servers as the Microsoft monopoly.
And speaking of Netcraft, Behlendorf gives them credit for opening doors for the Apache project. He contrasted the impossibility of knowing the current number of Linux desktop users with the statistics provided by Netcraft on server usage. Because Apache could claim — and demonstrate, thanks to Netcraft — a sizable, growing share of the pie, it got in the door at large corporations that might not have otherwise given it the time of day.
Bitkeeper vs. Subversion
A curmudgeon in the crowd asked Behlendorf to compare Bitkeeper and Subversion, the first being a proprietary program written by Larry McVoy, and the second a free software offering from CollabNet for source code version control.
Linus Torvalds and other Linux kernel hackers use Bitkeeper, much to the dismay of many free software purists. Subversion, according to its Web site, aims to become a “compelling replacement for CVS,” which has been widely used in the open source and free software camps for years.
After a brief technical comparison of the two — Subversion being more a centralized repository while Bitkeeper handles things as streams of patches — Brian said that he respects Larry McVoy, and he respects his right to code proprietary solutions.
But that hasn’t prevented McVoy from visiting the Subversion mailing list and threatening the developers with lawsuits if any of Bitkeeper’s features show up in Subversion. In fact, Behlendorf said, McVoy has altered the license on Bitkeeper so that nobody working on competitive applications can use it. As a result, he believes McVoy has turned a lot of people off from using Bitkeeper.
Sender ID licensing
Behlendorf said that sometimes geeks need to look up from the code and take note of events and politics in the real world. The licensing terms for Sender ID are one such thing. He gave Microsoft credit for at least letting the ITEF MARID working group know that it was filing a patent on Sender ID, but felt some of the licensing terms were too restrictive for open source projects.
One of the the things he objected to was conditional use: Microsoft says you are welcome to use the technology without charge today, but “we reserve the right to revoke it at any time.” And of course, there is all that bother about redistributing the code. It added up to a situation the Apache Software Foundation felt was untenable, thus the group issued a press release proclaiming it unusable by any of its projects.
Brian said offshoring is a fact of life, albeit an unpleasant one if you’ve lost your job. He said he bought a company in India to offshore some of CollabNet’s development work. He doesn’t begrudge other firms that do the same.
He also noted that within a week of Torvalds posting the message about the first version of Linux in an Internet newsgroup in 1991, he had received offers of help from around the globe, illustrating that open source is global in nature, not restricted by political boundaries.
The time flew by too quickly. Poor Brian never got a chance to even take a bite of his lunch, then had to leave immediately for another stop in Austin, at the University of Texas. On the other hand, he did keep a crowd of experienced software developers entertained and interested in what he was saying for an hour. If you get a chance to hear Behlendorf speak sometime, take advantage of it.