I'm spending the next three days in Orlando, Florida, attending ApacheCon 2000. The conference is being held at the Caribe Royale Resort Suites, which despite a strong conference turnout, is mainly inhabited by lots of parents and their young children, due to the proximity to Disney World. Orlando isn't high on my list of places to visit, especially during the prime spring-break travel season, but duty calls....
I missed the opening plenary session this morning, as my overnight travel plans had to be rearranged due to weather delays in San Francisco (imagine that!). But I heard from Ben Hyde, a member of the Apache Software Foundation, that I didn't miss much. Ken Coar, the conference technical chair, went over the conference schedule and layout, then the ASF members on the stage took questions from the floor. The questions were mostly predictable ones, about the time frame and features for the upcoming Apache 2.0 release. According to what I was told, the ASF has committed to having an alpha release out by the end of the conference (or next Monday at the latest).
I met Ben Laurie, co-author of O'Reilly's Apache: The Definitive Guide, for the first time this morning. He'd already been asked once when he was going to do a new edition of the book for Apache 2.0, and I predict that it won't be the last time he hears that question over the next three days. Now to answer the question: Ben will get started on a new edition of the book as soon as he feels that 2.0 is stable enough for him to write about, but we don't have any more definite information than that.
Walking back and forth from the conference center to my hotel room is a real visual experience. All the rooms, or suites to be precise, are in three towers that surround a courtyard with a huge crescent-shaped pool, complete with a waterfall and a 75-foot water slide. Crossing through the courtyard means seeing more exposed lily white body parts than I'd care to mention, interspersed with a few people who resemble cooked lobsters. Fortunately, the conference program is so packed with activities that I don't think many attendees will have the time to expose their delicate skin and risk sunburn.
There are several hundred people here for the conference. I bounced around from session to session today, so I can't report on the details of particular talks, but I can tell you what people were attending. At the first round of sessions this morning, it was standing room only for the first part of Stas Beckman's mod_perl tutorial. I don't know why his talk was in such a tiny room, but at least they were able to triple the room size for part two this afternoon. The session on the Jakarta Project was also overflowing. And I saw a good crowd at two talks on the business aspects of Open Source: "What is Open Source" and "The Cathedral Meets the Bazaar".
The conference even has a "night school" component, with sessions running from 6 to 7:30 today. Many attendees seemed content to forego having a full dinner and instead munch on some appetizers, in order to attend these sessions. There appeared to be lots of interest in Ryan Bloom's talk about migrating Apache 1.3 modules to Apache 2.0 and on Michael Meyer's presentation on the work his company, Advance Bank, has done to implement secure financial transactions with Open Source.
Perhaps part of the reason that people were willing to skip dinner is that there was a poolside dessert party this evening. Some people subscribe to the philosophy that "Life is uncertain; eat dessert first," but I've always thought that "Just eat dessert" is a better approach.
Now that I've come down from my sugar buzz, I think I'll call it a night.
The second day of ApacheCon 2000 started with a breakfast adventure. The Caribe Royale provides free breakfast to their guests--either a continental breakfast in the lobby of your building or a breakfast buffet in the main building. I made the mistake of going to the buffet. I felt like I had been transported back to summer camp, except that there were both parents and kids in the food line. The breakfast lineup included vats of slimy scrambled eggs, mystery meat, and crusty oatmeal. I think tomorrow I'll settle for the continental breakfast.
In my tour of the morning sessions, I noticed a sizeable crowd for Mark Wilcox's talk on Apache and LDAP. I had never thought much about how you might use Apache and LDAP together, but Mark discussed how the University of North Texas uses LDAP and presented compelling examples of using LDAP with Apache for access control (both user authentication and user authorization). He also presented some ideas about how LDAP might be used to help manage the configuration of Apache server farms.
In Jim Jagielski's talk on Web Hosting for Fame and Fortune, one issue that generated a far bit of discussion involved virtual hosting. The problem is figuring out how to cope when one of the sites you are hosting gets "slashdotted." Unfortunately, no one had any good solutions to this problem.
In the first keynote of the conference, Dr. Alfred Z. Spector, a Senior Technical Strategist for IBM Software and an Adjunct Professor for Columbia University Computer Science, presented his "Software Agenda". He started with a short discussion on IBM and Open Source, noting that "the world began with source" and commenting that "aggregative innovation" is one of the key benefits of Open Source. But the majority of his talk was on his concerns about the growing demand for software applications with ever increasing requirements. Spector's solution to the coming software bottleneck involves modularity and reusable abstractions, so that creating software becomes more like building a bridge. He feels that three things need to come together to make this happen:
Spector acknowledges that we are making progress in this area, with technologies like Enterprise JavaBeans, but feels that we still have a long way to go. And he feels that the Open Source community can play a role in helping to achieve his agenda.
The exhibit hall opened immediately after the keynote. The O'Reilly Network had a booth and seemed to have a steady stream of people registering at their terminals. Collab.Net was giving away cool bouncy balls with LEDs inside, and we all got treated to a juggling exhibition by Geoff Thorpe of C2Net Europe. And Border's Bookstore had a booth where they were selling lots of O'Reilly books, including Apache: The Definitive Guide and Writing Apache Modules with Perl and C, as well as books on Apache by other conference participants, such as Apache Server for Dummies, by Ken A. L. Coar, and Apache Server Unleashed, by Rich Bowen, Ken Coar, et. al.
For the first afternoon session, I attended a talk by Ryan Bloom on APR, the Apache Portable Runtime. Ryan explained that while Apache is part of the name, there's nothing that ties the APR to just Apache. APR is being designed as a portable runtime API for standard system calls across all supported platforms. APR was developed because Apache was becoming harder and harder to maintain across its various platforms. The first version of APR provides only those functions that are useful to a server application like Apache because of its importance to the Apache 2.0 effort. But future versions of APR may provide additional functions that are not required by Apache, and the expectation is that APR will be moved out of the Apache source tree and become a project in its own right.
Later in the afternoon, Theo Schlossnagle gave an excellent presentation on mod_backhand, a load balancing module for Apache. mod_backhand provides an alternate solution to round-robin DNS or hardware-based proxying, with some significant advantages over each of those techniques. mod_backhand's basic approach is to allow each machine in a cluster to proxy requests to any other machine in the cluster, thereby getting around the single point of failure of a hardware-based proxy solution. Even better, because mod_backhand is tied into Apache, it can collect resource utilization data and attempt to distribute requests based on that data, which is certainly better than the random distribution provided by round-robin DNS. Research is still ongoing in the area of pre-request resource allocation algorithms, but some people are installing mod_backhand just for resource utilization data.
After Brian Behlendorf's keynote on the State of the Foundation, there was a reception on the exhibit hall floor. Unlike at the dessert party the first night, there was beer and wine, which seemed to induce a more party-like atmosphere. There were also lots of waitpeople walking around with trays of various appetizers--an improvement over most receptions, where the food is placed on a couple of tables and the lines can get prohibitively long. In this case, you just had to spot the small clusters of people hovering around the popular appetizers. I spent most of the reception talking with James Duncan Davidson, who wrote the Java Servlet API and is now doing Java/XML work for Sun. We were at the opposite end of the room from the kitchen, so it took a while for the food trays to get to us, and from my scinetific observation I can tell you that the potstickers were the most popular item at the reception.
After the reception, there was another round of "night school" classes for the really dedicated attendees. I checked out part of Greg Stein's talk on WebDAV and Apache, which had a good crowd, despite the fact that it had already been a long day for most people. WebDAV (Distributed Authoring and Versioning) is a set of extensions to the HTTP protocol that promises to turn the Web into a writable medium. Greg covered the basics of WebDAV and also gave a brief overview of using the mod_dav module, which adds DAV capabilities to Apache.
It's been a long day. I think I'm going to go check out the swimming pool before calling it a night.
By getting up early this morning, I discovered something about the Caribe Royale: they turn off the waterfall that cascades into the pool each night and turn it back on again in the morning. The courtyard was blissfully quiet this morning, without the waterfall and the hordes of rambunctious children. But I'm glad that they turn the waterfall back on before most kids get up. Otherwise, dads would find it even easier to torment their kids like my dad did when I was five, when he had me convinced that they turn Niagara Falls off each night.
In my travels among the various conference sessions, I noticed a lot of interest in the three PHP talks that took place today. In the morning session, Nathan Wallace gave a talk entitled PHP: Hacker's Paradise. This talk was well attended and Nathan had quite a crowd around him after the talk, with people asking all sorts of technical questions about how to do various things with PHP. The other two talks, on using PHP with SQL backends and sessions and authentication, also seemed popular, despite the fact that they overlapped with talks on Apache 2.0 and the future of Apache beyond 2.0.
In another morning session, Mark Cox and Geoff Thorpe gave an interesting presentation on Apache E-Commerce Solutions. The buzzword-compatible title was a bit misleading, as the talk focussed primarily on Apache and SSL. Mark and Geoff talked a lot about the performance implications of using SSL, both in a single-server setup and with multiple servers that support a high-volume site. Their conclusion is that SSL session caching needs to be handled as a separate service, accessible to all of the web servers in a server farm, rather than being localized to each individual server.
In the final keynote of the conference, Patricia Sueltz, Sun Microsystem's President of Software Products and Platform, talked about three bets that Sun made in 1995: Internet, Network Services, and Bandwidth. They are making three new bets in 2000: Massive Scale, Network Stack, and Always On, Always Connected. Sueltz also talked about Sun's interaction with the open source community--how they are listening, learning, and responding. She also talked about her belief that a technology goes through four phases in its steward lifecycle: a nurturing phase, a community involvement phase, a standards bodies phase, and finally a public domain phase.
In the first afternoon session, Manoj Kasichainula gave a talk on the new features in Apache 2.0, including the new multiple-processing modules (MPMs). MPMs are modules that determine how requests are mapped to threads or processes, so that Apache can support different ways of using processes and threads, based on the underlying operating system and other factors. For example, when Apache is running on a Linux system, it is better to have multiple processes each running multiple threads that service requests. But on Windows, which doesn't handle multiple processes very well, the solution is one process with multiple threads. Most users don't have to worry about MPMs, as each instance of Apache uses a single MPM that is set at compile-time. The overall impetus for Apache 2.0 is improved scalability.
In the next session, a panel of Apache members held an interactive discussion about the future of Apache beyond 2.0. The panel included Ryan Bloom, Bill Stoddard, Greg Stein, Allan Edwards, Ken Coar, and Manoj Kasichainula. Each panelist talked some about what functionality he was interested in adding beyond Apache 2.0 and then the audience was given the opportunity to add their requests. The big issues seemed to be layered I/O and scalability.
Now that the conference is over, I'm headed back home to Colorado. It has been warm and sunny here in Orlando, in fact too warm for my tastes. But in less than 24 hours, I plan to be up in the mountains skiing, which is much more to my liking.
Return to the Apache DevCenter.
Copyright © 2009 O'Reilly Media, Inc.