I would be wary of deleting anything manually from the registry unless I knew exactly what the key/value combination did.
To understand the nightmare that is the registry, you first have to consider what it is and why it was created.
Generall, in UNIX and it's variants all of the real configuration is stored in /etc
This kind of central configuration makes everything easy to find and easy to manage, plus you only use what you need.
The registry was created back in the FAT days. Initially, all configuration for Windows apps was stored in .ini files. Because of the limitations of FAT, eventually all of these .ini files would accumulate and you would hit the maximum file limit and Windows would claim your hard drive was full, when in theory it wasn't. The registry was designed as a kind of 'central configuration'.
Microsoft made a couple of mistakes that still haunt us even in Windows Server 2003 and XP. The entire registry gets loaded into memory at boot up, regardless of whether or not you're using all of it. It's illogical. The registry makes no sense to the average user. Even with the creation of Windows Installer, which is supposed to keep track of everything installed, managing the registry is a nightmare. An uninstalled program never completely goes away.
For example, there was in Windows 2000 a POSIX subsystem. According to MS, it is no longer present in XP -- but it's registry keys are. Even Microsoft has problems managing the registry.
The best rule of thumb is to reinstall and use only what you need. Don't install and uninstall.
hardware
prices 
