Posted on March 22, 1999 at 12:41:36 AM Central.
MacOS X Server was originally announced at January's MacWorld Expo with a price of $999. Many complaints later, Apple shipped it with a list price of $499, and an educational version for $250. Developers can get a 5-user version for $99; combined with the $89 student developer program, students can get MacOS X Server for under $200. Apple is also shipping G3 "servers" with the OS preinstalled; these are just blue & white Power Mac G3 models with SCSI and a quad-Ethernet card (presumably for NetBooting).
Since Apple's new server OS is available to many people at a relatively low price, there is some confusion about who would and wouldn't benefit from using this new, buzzword-compliant system. Apple has been narrowly targeting MacOS X Server at Web serving, publishing, and education; this is partly a matter of marketing but is also intended to protect Mac users. MacOS X Server is not as easy to use as MacOS 8, requires more RAM and disk space, supports few peripherals, and doesn't run all MacOS 8 applications. Thus I don't consider it an "upgrade" from MacOS 8; it's a different operating system that happens to look similar and run on the same hardware. Speaking of hardware, MacOS X Server is only "supported" on desktop G3 Macintoshes (I put supported in quotes because, to my knowledge, Apple doesn't offer any meaningful tech support anyway). However, the installer will install it on any PCI Mac. Whether it runs correctly or not, I don't know.
As a server for a Mac workgroup, MacOS X Server fits the bill perfectly. I think MacOS X Server should also be useful for software developers, especially those who prefer to work in a Unix environment. It is probably a must-have for WebObjects developers, but I don't have any personal experience with WebObjects. There doesn't seem to be anything that recommends it for other uses.
Contrary to some reports, MacOS X Server does include the Blue Box compatibility environment that allows MacOS 8.5.1 to run as an application, and thus supports running classic MacOS applications. Apple is strongly urging users not to run the Blue Box on servers, saying that it degrades performance. These statements leave many people who want to use the new OS as a workstation scratching their heads. I have yet to use MacOS X Server hands-on, but reports that I've read say that some applications run slower and some run faster under the Blue Box. I assume that graphics apps and games are slowed down by the lack of graphic acelleration, while I/O-intensive apps run faster because of better disk caching and virtual memory. Apple's claim that other applications slow down when the Blue Box is running in the background is not very interesting; if I wanted to run a Yellow Box application I could just shut down the Blue Box first.
Apple is claiming that a G3 server running MacOS X Server is the fastest Apache Web server under $5,000; the press release included a specific comparison against NT, Intel/Linux, and SPARC/Solaris systems. I don't consider this a fair benchmark, since NT includes IIS and Solaris includes Sun WebServer, both of which are faster than Apache in many tests. A few users may specifically need Apache, but if you're only serving files it's probably smartest to use the Web server that is optimized for your platform.
Although I've been hearing about Network Computers for a few years, Apple is the first company to deliver a sensible NC solution. The reason why I don't consider solutions from Sun and NCI to be sensible is that they don't run any mainstream applications; they require applications to be written in Java, which they have relatively slow support for. In contrast, an NC solution using iMacs and MacOS X Server runs nearly all MacOS applications, uses the familar Mac interface, etc. It can do this because it simply loads MacOS 8.5 over the network and runs it as usual. Accounts are managed through Apple's NetInfo directory service which allows users to log in to any iMac in a domain for which they have an account. NetBoot is a possible explanation for the five Fast Ethernet ports that come standard on Apple's G3 servers; sending copies of MacOS over the network requires a lot of bandwidth.
One of the big surprises of Apple's MacOS X Server announcement was the news that they have released the source code to large parts of the operating system under the name "Darwin". Although the names are different, Darwin is the exact same code that MacOS X Server is built from. The important difference is that many parts are missing, including Apple's famous GUI. Apple claims that Darwin is a complete OS, but it appears that some important drivers may be missing from the source distribution. Therefore I am not sure whether Darwin can stand alone or whether it still requires parts from MacOS X Server. Apple is planning on releasing a pre-compiled distribution of Darwin which should settle the issue.
There has been debate in the Open Source community about Apple's new Apple Public Source License. Eric Raymond of the Open Source Initiative, who had a hand in the drafting of this new license, says that it qualifies as Open Source despite its various restrictions. Software in the Public Interest's Bruce Perens says that he doesn't consider the license to be Open Source. Since the exact ownership of the Open Source trademark is in dispute, I'll leave it up to you to decide who to listen to. Richard Stallman also voiced his concerns about how the license affects the free software community.
Perhaps a bigger mystery than the licensing is who will actually use Darwin. While MacOS X Server includes a lot of cool technology, almost all of it was left out of Darwin, since Apple considers it to be a competetive advantage. MacOS X Server developers will probably find the Darwin code useful for tracking down bugs. LinuxPPC users are unlikely to switch, since Linux has more features than Darwin (notably a GUI), supports more applications, and is faster. I expect that most MkLinux users will switch to Darwin because development on MkLinux has slowed to a crawl in recent months. Theoretically it would be possible to build a new, completely Open Source OS by combining Darwin with GNUstep and other existing pieces of Open Source software, but I don't see how the result would have any advantage over Linux.
A possible side effect of the Darwin source release is that the source code contains many previously-unknown low-level details on Apple hardware, which could theoretically be used to improve hardware support on other OSes. However, LinuxPPC already runs on most Power Macs and Be refuses to use the information.
I am tempted to compare MacOS X Server and BeOS, but this is pointless. Apple has assured that there are no computers on which both BeOS and MacOS X Server are supported, making a fair comparison impossible. I may be able to run both operating systems on my 7600, but after a comparison the loser would invariably complain that testing their software on such an old machine is unfair. Apple would perfer that I use its OS on a new translucent Mac, while BeOS shines on dual Pentium III systems.