Why does everyone show a different list of Windows versions when making that tired old quote about every other version being good (or bad)?
Perhaps because if you use the actual list, it's not so clear cut.
First off, there are the major DOS-based versions:
1, 2, 3, 95, 98, ME
But that's not quite right, because there are minor versions with significant updates, and splits within. So you've really got:
1, 2, 2.1(/286), 2.1(/386 -- first release with protected mode), 3, 3.1, 3.11, 3.2 (Chinese), 3.1x/3.2 with Win32 (32-bit extensions), 95, 95 SP1, 95 OSR1 (first release with IE bundled), 95 OSR2, 95 OSR2 USB Supplement, 95 OSR2.1, 95 OSR2.5, 98, 98SE, ME.
Then in (pseudo-)parallel development, you have the NT versions:
3*, 4, 2000, XP, Vista, 7, 8, 10
But then there are the minor versions of those:
3.1*, 3.5, 3.51, 4.0, 2000, 2000 SP1-4, XP, XP SP1-SP3, Vista, Vista SP1-SP2, 7, 7SP1, 8, 8.1, 10
* The first version of NT was 3.1, to keep the numbers parallel with the DOS-based version of Windows the GUI was from.
Finally, interleaving those based on release date, you have (NT in bold)
2.1(/386 -- first release with protected mode)
Windows for Workgroups 3.1,3.11
3.1x/3.2 with Win32s (32-bit extensions halfway compatible with NT)
95 OSR1 (first release with IE bundled)
95 OSR2 USB Supplement
Now, I've taken some liberties with this list, not including most x.0x releases because they generally did not include much new functionality, but NT 3.51 and all of the 95 OSRs did include some major functionality changes. Although the later NT version Service Packs often did include significant functionality changes (most notably XP SP3 which made XP somewhat secure, and Vista SP1, which made Vista actually work) but since at that point there weren't parallel versions, I combined the SPs into a single entry because that makes them a single cumulative set of functionality updates.
So even this list is debatable, but it's a damned sight closer than any other I've seen. Still, it shows how absurd the alternating quality hypothesis is -- not because you can't come up with a list in which the odd versions are good, but because there are so many ways to crop and shuffle this list that you can just as easily come up with an equally definitive list which proves that the even versions are the good ones, or the prime-indexed ones, or whichever particular pattern you'd prefer.