Looks like you get your wish trvlr.
I'd stayed out of this as what the actual code is (and does) in the 1st Sector of a drive is still partly a mystery to me. The MS MBR does only look for an ACTIVE partition's boot sector. The boot sector actually locates the boot loader file/s within the file system (some how - it is extremely difficult to try and compare MS sector counts with the drive's geometry, and the BIOS LBA handling, and the INT13 calls and the little endian notation screws this around also) . Yes, there is only one NTLDR used to boot and it is located in the Active partition. What NTLDR actually does, in detail, is not publicly known (or not widely).
What follows is from using Microsoft's DiskProbe.
One of my HDD's has 2 partitions, primary and extended. There are 6 drive letters, only one of which contains an OS. That drive's (the OS drive) Sector 0 starts out with the EB 5B and sure enough, there is a FA instruction at 5B bytes from there (offset 5D). And the text at the end would seem to indicate that it is a boot sector:
A disk read error occurred...A kernel file is missing from the disk...A kernel file is too discontiguous... Insert a system diskette and restart the system... NTLDR is compressed.
Format is NTFS, not that it matters. A FAT data drive has the same messages. While another FAT data drive has:
BOOT: Couldn't find NTLDR. BOOT: I/O error reading disk...Please insert another disk.NTLDR
Another drive has:
Invalid system disk...Disk I/O error... Replace the disk, and then press any key IOSYSMSDOS SYS...WINBOOT SYS. This is a FAT32 data drive.
All the drives have the 55 AA End of Sector marker, of course. I am fairly sure that I am reading the correct sectors as the BPB (Bios Paramater Block) does change.
To see if there was a correlation between OS and boot sectors. I moved one of my W2k OS's from drive to drive. Formatted the old drive, installed XP on it. The drive Sector 0 didn't look to have changed. Which only proves that all NT type OS's MAY have the same boot sector (although I remember reading about some change MS had made). It could also be that the MS formating programs automatically put a boot sector down on Sector 0 during the format process. I don't know.
NOW, the reason I did respond: softcrypt
Re your #3: Don't confuse the Microsoft MBR code with anything else that uses physical disk sector 0 to boot an OS. A net search will provide lots of "build your own boot loader code". Linux's LILO is actually part one of the kernel. It does rewrite the MBR. So the statement that you and I both read somewhere "the MBR is OS independent" is only correct for the Microsoft OSs. Here is a cut and paste from one of the things I have on the LILO boot loader:
This MBR code is only the 'first stage' of the complete LILO program. You can find documented sources by LILO's author, Werner Almesberger, on most Linux archive sites. Try looking here first: sunsite.unc.edu/pub/Linux/system/boot/lilo/ for the archive, lilo-21.tar.gz or higher. The code for the MBR is found in the file, first.S, whereas the much larger portion of the Assembly code is found in the file, second.S.
I have the file, just haven't reinstalled a real Linux that can uncompress it yet (one downfall of the Knoppix CD). I have a copy of a brief discussion on the actual code, but it is nothing to write home about.
For your #1: you are right, nothing is changed in the MBR. Take a look at the code on dougknox.com if you want the quickest way of making a bootsect.dos file.
For your #2: Take a look at the boot sequence and you will see why there is that difference between the 9x and NT boot sectors. 9x calls for the IO.SYS and MSDOS.SYS files. NT type OS's call for NTLDR. It is actually the BIOS and NTLDR that had the 1024 limitation - and the 7.8GB limitation.