Globalization and women, copyright infringement in open source, and other news from LinuxWorld

   Print.Print
Email.Email weblog link
Blog this.Blog this

Andy Oram
Jan. 26, 2004 07:24 AM
Permalink

Atom feed for this author. RSS 1.0 feed for this author. RSS 2.0 feed for this author.

LinuxWorld in January 2004 looked mostly like more of what was at last year's LinuxWorlds. No radical departures; just more serious talk of moving to the desktop, more and faster clusters, more products aimed at ERM and other back office services, more advances in the tools available to support Java. But more is better. Here I'll report some conversations that gave me new perspectives and some companies I met with.

How globalization could promote the role of women in computing

At LinuxWorld I met Margo Seltzer, who invented the Berkeley Database (BDB) at UC Berkeley and now shares her time between Sleepycat Software (the company commercializing BDB, where she is CTO) and Harvard University, where she is Associate Dean for Computer Science and Engineering.

Seltzer told me how BSD implementor Keith Bostic took assessment of BSDI many years ago and found that its communications and operations were lacking. An early example of a free software development organization, BSDI depended on communication and sharing among a far-flung set of programmers, but the communication and sharing weren't happening. Bostic realized these virtues were traditionally associated with women, and started hiring with that in mind; ending up with a programming staff that was half women. Communication and coordination improved dramatically, and so did results.

Seltzer worked with Bostic at BSDI and now works with him at Sleepycat Software. Wherever she has been a manager, she promotes personnel policies that make it easy for people to keep jobs while raising children. These are fairly familiar adaptations that apply to a wide range of professions: flex time, telecommuting, part-time opportunities, and making children feel comfortable in the workplace. She says this flexibility has allowed her to employ a lot of highly qualified and highly dedicated women, including one who is "one of the best engineers she's ever known."

What can the computer field do to attract more women? Seltzer says that the process has to start early in life, long before the job interview. Schools, universities, and even game manufacturers (perhaps especially game manufacturers!) should adjust their offerings to appeal to girls. Seltzer says the computer science field is already moving in that direction, a trend I can confirm from my reading of the Communications of the ACM over the decades. Computer science programs are broader now than they used to be, covering not only the technology, but the interaction between technology and practical domains where the technology is applied.

There is a well-established tendency (not true for every individual, of course) for men to show interest in engineering and technology for its own sake, while women are more often among those who see engineering and technology as means to an end, and who want to see a useful benefit at the end of the line.

Furthermore, men often like clearly delimited tasks with fixed goals and requirements, while women are often better at doing systems analysis and fitting solutions to real-world problems. Microsoft is one company where Seltzer says she sees the different genders ending up categorized that way.

Now, what jobs are easiest to outsource overseas? Precisely the delimited tasks with fixed goals and requirements. The job of systems analysis and fitting solutions to real-world problems remains in the controlling country. Therefore, Seltzer believes that the movement overseas raises the profile and status of women in the computing industry.

Another female colleague I know who runs her own consulting company told me this week that open source has many qualities we traditionally think of as feminine. It involves communication, cooperation, and a willingness to put together contributions from many people without worrying who gets credit. Furthermore, nothing is ever finished and neat; the process is ongoing.

After talking to Margo Seltzer, I went to a party and chatted with Perl porter Kurt Starsinic, who works for the famous financial information and services firm Bloomberg. He describes them as "one of the best places to work in the world." They try to find top-notch personnel and then largely trust them to work independently--but they are expected to find a peer to audit their coding and related practices. Good communicators are demanded there, and the very architectural layout of walls and staircases ensures frequent interaction. In consequence, comraderie and interaction are at a high level. I also saw a suggestion of their emphasis on the quality and nurturing of employees at their website.

Hearing all this from Starsinic, I asked him on impulse, "How many women programmers does Bloomberg have?" He answered, "A lot!"

Even if companies adapt, women can expect sniping from the sidelines. An article in the Wednesday New York times reports grumbling from a conservative group at Princeton (which has a female president) because four of the university's senior positions are occupied by men and four by women. The grumblers, staunch preservers of the social order, can't believe that the female appointees could actually have won their places through their qualifications. Someone should hand these critics the statistics that show how women are already in the majority in college enrollment and among entrants to a number of professions. I also heard at LinuxWorld that a United Nations agency is raising the social and economic status of women in Africa by training them to be computer system administrators.

Readers interested in learning about and working on such issues can consult:

Who are the copyright infringers in open source?

On a tip from a colleague, I stopped by Black Duck Software to see their system for tracking violations of open source licenses. This is no mere diversion for embittered hackers; it's a critical procedure for large companies who have reason to worry that their programmers have violated licenses.

Black Duck CEO Douglas Levin cited, for instance, a case reported by investment banker Broadview. When IBM acquired Think Dynamics, a painstaking manual examination of its code revealed 80 to 100 examples of open source code that Think Dynamics programmers had passed off as their own. As a result, the price of that company went down from 67 million dollars to 46 million--not a happy moment for its owners and shareholders, I'm sure.

Readers cannot miss, I'm sure, the irony that this very IBM, which was so diligent in tracking down license violations, is being accused of the same by SCO Group. Indeed, the absurdity of SCO's suit should be obvious, because who would try to hide infringing material in open source code? Obviously, the true danger lies in the other direction: code generously donated to the world by free software programmers is being hidden in binary files and used to pump up the profits of proprietary software vendors.

Black Duck staff are not the only ones worked up over this scandal. I talked to a leading open source software developer during the week who said his business was virtually ruined by competitors who incorporated his code into their products, stripping off his GPL notices and sometimes distributing it in binary fashion. It's very hard to find infringement in binary files; in one case he did so because the guilty programmers were too lazy to change his unique error messages.

It's also hard to document cases of copyleft infringment, except in rare instances such as Think Dynamics. Often the aggrieved programmer lacks the resources to sue a large and unscrupulous company, so there's no established, publicized outcome. On the other hand, when a programmer persuades a vendor to admit the infringement and pay up, it would be unfair to publicize the issue and punish the vendor for ultimately doing the right thing. However, Levin told me that many companies have invested large amounts of money in proprietary software products, only to find at the last minute that they had to release everything under an open license because somebody discovered free software at its base.

But Black Duck is counting on there being enough honest software vendors, and enough settlements with real penalties, to drive sales for their very interesting product. The company tracks dozens of well-known sites for open-source software. It takes bits of code from source files throughout the software and creates fingerprints that it stores in a database provided to its customers. It also provides a snazzy front end, an Eclipse plug-in, that allows programmers and managers to quickly scan their own source files, log instances of code found from open-source projects, and determine what license covers the code so they can deal with it properly.

Volunteerism alive and well

We all welcome investment by big corporations in free software. But I found LinuxWorld's Org pavilion, representing largely volunteer groups and non-profits (some of course with corporate backing), to be the most crowded in the show. OpenOffice.org and the Linux Terminal Server Project, located side by side, were positively a fire hazard.

I lunched with Jim McQuillan, founder and leader of LTSP, who does consulting for health care organizations in his spare time. He told me that HIPAA is a big boon to consultants whose Y2K business ended. HIPAA is the well-known and controversial privacy regulations that you've probably run into because your local clinic and hospital have required you to sign forms concerning your privacy rights. HIPAA is largely a beneficial regulation passed in the Clinton era, weakened as one would expect by the Bush administration, but extremely subtle and complex in its implementation.

Of course, the popularity of the Org pavilion was a bit exaggerated by the tendency of non-profits to rent smaller spaces than commercial vendors. But the bustle in that area, along with the interesting projects shown off there, remind us that the spirit of personal contribution remains high in the Linux arena--particularly among New York user groups and organization representatives, who deserve a big collective thank-you for the time they put into the conference.

One of the non-profits I visited is the Free Standards Group, who have launched an effort to promote free desktop software that is accessible to the blind, those unable to type or use a mouse, and other disabled users.

A cluster of related vendors

I don't think any single application dominated LinuxWorld as much as clustering. You couldn't cast an eye about anywhere without noticing companies with key products in the clustering area, and even more companies (in fact, nearly every company) employed clusters as part of a solution to some more specific need.

Some vendors I talked to are:

  • Donald Becker, one of the people who could be said to have started it all. The company he founded, Scyld Computing Corporation continues to release new Beowulf-based products after its purchase by Penguin Computing, and now offers both Itanium and Opteron clusters.

  • Qlusters has up-and-coming clustering software.

  • Chuck Foley of InfiniCon sees three trends coming together to create a new era of high-performance computing: Linux clustering, commodity 64-bit chips, and InfiniBand, which offers up to 30-Gigabit networking. One of InfiniCon's customers, Pennsylvania State University, recently clocked in with the world's 89th fastest supercomputer (1.3 teraflops using 160 2-processor, 32-bit nodes). The previous cluster ranked number 89 required 1100 nodes to get even close to that performance, showing the price/performance improvements that are currently being achieved in Linux clustering. With 64-bit processors, another InfiniCon customer hopes to make it up to 5th in the next world supercomputing listing.

  • A staff person at RLX Technologies told me they started offering their clusters for high availability (that is, providing reliable, mission-critical services), but have increasingly found them being used also for high performance (that is, computations in well-known domains like seismology and bioinformatics). They provide turn-key solutions integrating Linux and management software with the processors and either Infiniband or FiberChannel.

  • SteelEye makes software for monitoring and doing automatic restarts or failovers for clusters; it is heavily resold by IBM and HP. They feature fine-grained monitoring: for instance, if a system fails to respond because the database went down, the monitor can simply restart the database instead of failing the whole system.

  • Pogo Linux offers a range of systems, focusing on servers and storage RAID-based systems.

  • A number of companies with multiprocessor-related compiler tool chains were hanging out in the AMD area. Etnus offers some sophisticated technology, including a multithreaded debugger, for the AMD Opteron and a number of other Unix and Linux platforms. The Portland Group is another company demoing a very cool-looking visual debugger that can show you what each thread is doing and step through each thread independently. They will soon release a profiler as well. PathScale concentrates on Opteron support, intensively examining every opportunity to squeeze performance improvements from that chip. They claim the best performance of any vendor on the Opteron, and say that they do equally well speeding up integer and floating-point code. Furthermore, their commands are compatible with gcc, allowing use with the same scripts and makefiles.

Some larger vendors

For reasons that may well be totally arbitrary, I got more contacts with IBM than with any other large company at LinuxWorld this year, and therefore heard more of what they had to say. IBM has ported Linux to its 64-bit POWER systems, the latest in its RISC-based genus of which the PowerPC chip is the most famous species. A number of ISVs have already ported their applications to Linux on 64-bit POWER; SAP is in the works.

VP Brian Connors stressed that the new POWER prices were comparable to the Xeon, not the Itanium: "64-bit systems at 32-bit prices." Both an IBM compiler and the free gcc are being enhanced to support the processor. In the larger scheme of competition, Connors thought it important for Linux to be a viable option on many different types of processors: "The architecture should be open at every layer."

VP Adam Jollans asserted, "As we go into this year, Linux will enter the space of ERP and other big enterprise applications." This can be credited partly to the new 2.6 kernel, which scales up better than the 2.4 one.

When I asked Jollans what IBM was doing to improve the experience on desktop Linux, he noted that a number of Lotus clients can be run through a Web browser to allow Linux access. (Lotus server applications have already been ported to Linux.) He reiterated the strategy for moving Linux gradually to the desktop that had been articulated by IBM and others at the Desktop Linux Conference.

IBM thinks it has found a way to capitalize on the well-publicized resistance Microsoft customers are showing in the face of its upgrade strategy. When security upgrades vanish from Windows NT, IBM is hoping to get the system's two-million-strong customer base to move to Linux instead of throwing away their hardware and buying new Windows XP machines.

By the way, if Microsoft is trembling at the edge of a chasm, it would take quite a while to notify all the customers that continue to show it loyalty, and probably even harder to convince them of that fact. On Friday, the company reported better than expected results for the quarter, and the highest revenue ever.

MySQL AB has quite a brood of followers in their conference area, a little conference within a conference that grows year to year; this year even Intel had representatives in the MySQL area. (An actual MySQL conference took place last year and will be held again this coming April.) I have reported on some of these companies in my LinuxWorld weblog from last August.

One company I noticed there for the first time was High-Availability.Com, which makes a cluster to facilitate MySQL replication and resiliency. It supports a wide range of operating systems. Arkeia was also present, showing their tape backup system.

We have all heard that PHP decided to bundle SQLite with their free distribution instead of MySQL. A volunteer at the New York PHP booth told me there was no single show-stopper that decided against MySQL, just a grab-bag of licensing worries. But Zend Technologies Ltd., the company responsible for designing new versions of PHP and creating its core engine, licenses MySQL and bundles it with their products, which include an IDE and a highly regarded performance enhancer.

Doron Gerstel, CEO of Zend, told me that about 50% of PHP sites use MySQL as the back-end; another 25% use Oracle and the rest are divided among other databases. Zend is seeing PHP increasingly in critical applications such as e-ticketing. Eventually, they plan to provide PHP with a query API that eliminates the need for SQL.

Other companies I happened upon

With all the crush of clustering and server companies, hardly anything in other areas--such as wireless devices or embedded systems--made an appearance at LinuxWorld. One company that combined both those traits was AML, which makes hand-held terminals and other small devices meant for data collection in factories and other facilities. My question: what country will be the first to send a Linux system to Mars?

One of the trends I've noticed (particularly when reading Sys Admin magazine) involves tools that take automation to the network. Whereas we've had tools since time immemorial that check for intruders or monitor performance on a single system, the new crop of tools do the same thing for all the servers or all the nodes you're responsible for, and combine results.

One of the companies representing this trend is Net Integration Technologies, which has been marketing its products since 2000. Suppose you install a new network interface on a router. You'd like all the clients behind the router to know they can use this interface; Net Integration's tool configures them automatically to do so. It also allows quick disaster recovery, maintaining a great deal of state about the system. Representative Dan Wensley told me they consider their product well-suited to small and medium businesses, and hope it helps push Linux into that market.

I stopped by EmperorLinux, which sells a variety of laptops and notebooks with Linux preinstalled, along with technical support.

Well, that wraps up what I as a single person experienced at the last LinuxWorld to be held in New York. I couldn't be everywhere and hear everything (in fact, by the end I didn't feel like I was anywhere and could hardly hear anything) so I know there were many interesting announcements for which you'll have to read other reports.

Andy Oram is an editor for O'Reilly Media, specializing in Linux and free software books, and a member of Computer Professionals for Social Responsibility. His web site is www.praxagora.com/andyo.