Apache Week
   
   Issue 157, 16th April 1999:  

Copyright ©2020 Red Hat, Inc

In this issue


Apache Status

Apache Site: www.apache.org
Release: 1.3.6 (Released 25th March 1999) (local download sites)
Beta: None

Apache 1.3.6 is the current stable release. Users of Apache 1.3.4 and earlier on Unix systems should upgrade to this version. Read the Guide to 1.3.6 for information about changes between 1.3.4 and 1.3.6 and between 1.2 and 1.3.6.

Most bugs listed below include a link to the entry in the Apache bug database where the problem is being tracked. These entries are called "PR"s (Problem Reports). Some bugs do not correspond to problem reports if they are found by developers.

Bugs in 1.3.6

These bugs have been found in 1.3.6 and will be fixed in the next release.

Because of the major differences between Windows and Unix, these are separated into bugs which affect Windows systems only, and other bugs (which may affect Windows as well). Unix users can ignore the bugs listed in the Windows section.

Windows-specific Bugs

  • htpasswd.exe was only using the first eight characters of the password.


Under Development

Patches for bugs in Apache 1.3.6 will be made available in the apply_to_1.3.6 subdirectory of the patches directory on the Apache site. Some new features and other unofficial patches are available in the 1.3 patches directory. For details of all previously reported bugs, see the Apache bug database and known bugs pages. Many common configuration questions are answered in the Apache FAQ.

Per-directory environment variables

The SetEnv directive can be used to set environment variables which are then available to CGI programs and SSIs. This directive is currently only available in the main server configuration and virtual hosts, but not in containers such as <Directory> or in .htaccess files. From the next release of Apache, SetEnv, UnSetEnv and PassEnv will all be available for use in per-directory locations as well.


Biased Performance Tests Challenged

MindCraft have recently released a report comparing the performance of Apache on Linux against IIS running on NT. They also tested Samba (for file sharing) on Linux against NT's built-in file sharing. Their summary of the results is: Microsoft Windows NT Server 4.0 is 2.5 times faster than Linux as a File Server and 3.7 times faster as a Web Server.

This result is significantly different to other performance tests. For example, ZDNET's own tests showed Samba beating NT in The Best Windows File Server: Linux! and Samba 2.0: A License To Kill NT?, and Linux beating NT for web serving in Linux Is The Web Server's Choice. Looking at the MindCraft tests in detail, it is clear that they compared a heavily tuned and optimised NT system against a sub-optimal Linux configuration. It is also notable that the MindCraft test was paid for by Microsoft, and that MindCraft state: "With our custom performance testing service, we work with you to define test goals. Then we put together the necessary tools and do the testing. We report the results back to you in a form that satisfies the test goals".

It is usually always possible to come up with performance tests that show one particular application out-performing another, because applications can be tuned to perform will in a test environment. It is not surprising that a test can be found which shows IIS having better performance than Apache, since Apache concentrates on getting good performance in real world situations. In contrast, servers optimised for tests will give good performance in situations where there are very quick low-latency connections to clients.

Even allowing for the performance difference expected because the tests do not simulate the real use of a server, there are significant flaws in the test environment that make these results unreliable. A number of reactions to the report have been produced: A look at the Mindcraft report and Mindcraft Reality Check both contain detailed and reasoned arguments. ZDNet has reported on the controvesy in NT beats Linux ... maybe.

MindCraft performed expert tuning on their NT system to get the best performance. They extensively configured NT and IIS, for example, by modifing the NT registry to get best performance against local network clients. By contrast, they picked a Linux kernel version which had known performance problems when used against Windows clients. They also used a RAID disk controller which is not fully supported by Linux. In Samba, they turned on an option which was unnecessary, but results in significantly lower performance. These are just some examples of the ways in which the tests were heavily weighted in favour of NT.

It is clear from the MindCraft report that they have designed a complete environment where NT was bound to win the performance tests. This cannot be described as unbiased. If MindCraft were interested in publishing accurate unbiased results they should perform these tests again, optimising the Linux, Samba and Apache hardware and software in they same way they optimised the IIS and NT hardware and software in their test.