Microsoft's second version of Windows 95 included a new disk format, FAT32. FAT32 broke the 2-GB disk size barrier -- finally, Windows could recognize larger volumes. It also used disk space more efficiently than the older FAT16 when storing small files.
Microsoft used a different file system for Windows NT: the NT File System (NTFS). NTFS eliminated the 2-GB barrier too, but also included many features unnecessary in Windows 95, including file and directory security and permissions. Of course, NT couldn't read or write FAT32 (without third-party add-ons) and Windows 95 couldn't read or write NTFS (again, without add-ons).
This resulted in some unique storage arrangements for systems that needed to be set up for dual boot, but also required disks larger than 2 GB. A computer could have some disks available to only one of its operating systems and some available only to the other, with very little common space. It was often an interesting adventure to get a large file from a FAT32 partition to an NTFS partition on the same computer.
Recently, Microsoft released the long-awaited Windows 2000, which continues to support the NTFS -- indeed, some features of Windows 2000's Active Directory require an NTFS partition. But Windows 2000 also has built-in support for FAT32.
Why is that a big deal? Most servers run a single OS, and don't need to dual boot. Why worry about FAT16 or FAT32?
Because if you format your primary NT boot partition as NTFS, you could end up in a world of hurt.
Consider this scenario: An incorrect device driver is installed on an NT system. The driver causes the system to fail with a blue screen within seconds of any login attempt. If the boot partition is FAT16 or FAT32, you can boot with Windows 98's FAT32 Emergency Boot Disk and fix the driver. If the partition is NTFS, you may have problems. Windows 2000 does have more emergency boot options than NT does, including safe mode and a boot disk set, but I haven't had enough experience with those features to put my fate in their hands.
Of course, there are safety measures you can take. My favorite is to install two copies of NT or 2000 on the boot partition. (I normally keep both copies on the same drive, but it's better to put them on different drives.) One is the normal OS that runs day in and day out. The other lives in a different directory and remains pristine. When the first copy of NT becomes corrupt -- note the use of when, not if -- I can boot to the other copy and make repairs without resorting to drastic measures.
If you use this technique, having an NTFS boot disk isn't a big risk. But I still think keeping the boot drive a FAT partition makes sense. Suppose, for example, you have a server that runs normally. No administrator logs on for months at a time. Then you go to log in and -- ouch -- you can't remember the password for the system administrator ID. No problem -- you simply bounce to the alternate version of NT, where the administrator ID is still "administrator" and has no password, by rebooting and selecting it from the BOOT.INI boot menu. You're in.
The permissions on the NTFS disk don't allow this access.
Wouldn't it be much easier to stick a diskette in the drive, boot off that single floppy, get access to the C: disk, and fix the problem? By the way, if you are ever stuck without a password, rename the SAM file from the registry directory and reboot. A new SAM will be created with the "built-in" user IDs. Make any other necessary fixes, then boot again using the diskette. Move the old SAM back and reboot. Voil