Re. Netscape, IE on Windows surely helped, but Netscape did the no. 1 mistake and tried to rewrite everything from scratch - I can not think of any working product where it is better to throw everything out and redo everything "the right way", rather than iteratively improving the code. It's knowingly throwing out all business knowledge improvement and bugfixing you have ever done. It killed them, and would have even if IE was never bundled with Windows. In contrast, Windows 10 and Office 2016 contains code that can be traced more than 25 years back, and as evident with Vista development, MS is not afraid to accept when something has gone horribly wrong and throw it out and try again. Note that is not the same as rewriting, it is accepting that you spend a lot of resources making a prototype that ultimately showed no promise.
Not that MS does not do terrible things as a business. They do, exactly as any other business with that size. They have to or they will cease to exist in a few years.
This emphasizes a restriction technical debt may place on your company if your starting foundation is sufficiently poor. Lotus did a combination of these two fatal mistakes, they had a small team of 7 or so guys write the first version in 8086/8 assembly language, and I'm not sure exactly what happened to that team, but their EMS memory (https://en.wikipedia.org/wiki/Expanded_memory) mid-life kicker was unauthorized work done by a couple of guys who had left the company by the time Lotus used it in desperation because their complete rewrite of the code in C was failing so badly.
For that matter, their first Macintosh program, Jazz (https://en.wikipedia.org/wiki/Lotus_Jazz) was fatally flawed in development. They again wrote it in assembly "for efficiency", but without close communications between team members. Instead, every N days each would bring a floppy disk with their updated code and a merge would be done. This resulted in a lot of duplication of functionality and a bloated binary (hitting the platform where it was at its worst); which among other things resulted in user frustration at having to swap floppy disks all the time as the Wikipedia article mentions.
Apple is a partial counterexample: Jobs rode a small team hard in creating the first rather limited assembly OS, software wise they largely wandered around in the wilderness and failed the complete rewrite (https://en.wikipedia.org/wiki/Copland_(operating_system)), then solved their problem by buying NeXT (and, oh, started bringing Jobs back into the fold, one of the very few non-programmers who can manage them), and doing a wholesale replacement with NeXTSTEP.
Not that MS does not do terrible things as a business. They do, exactly as any other business with that size. They have to or they will cease to exist in a few years.
Disagree. Microsoft's terrible things included screwing pretty much everybody who did business with them at the peer level, "You made a mistake, you trusted us," said 3Com founder Bob Metcalfe, quoting an unnamed Microsoft executive." (http://www.pbs.org/cringely/pulpit/2002/pulpit_20020620_0007...). You do enough of this and you run out of healthy companies willing to do business with you, and life becomes prohibitively expensive. Which phone carrier is willing to partner with Microsoft after e.g. the Microsoft Kin: https://en.wikipedia.org/wiki/Microsoft_Kin and I'm sure other stunts?
I can't think of another company at this level like this. Apple and Oracle, like Microsoft companies built on the vision of a single man, share some characteristics, but Apple hasn't screwed over every supplier, and works well enough with phone carriers. Has Oracle tried to screw over big iron competitors after it bought Sun?
The official Windows phone unit was doing its best to kill the Kin, it had to switch to Windows CE, it was required to support Exchange, and a lot of the code supplied to the Kin group, especially for Exchange, was in horrible shape.
Ah, and another example of this sort of bad Microsoft behavior: When they bought Danger, they ignored its contractual responsibilities to T-Mobile and expected to immediately repurpose Danger's technical employees. Oops, T-Mobile refused to put up with that, and those employees had to spend some time developing the last Sidekick models (and all this had a morale crushing effect, above and beyond that of being absorbed by Microsoft with its stack ranking el. al.).
IIRC, Microsoft were using hidden APIs to accomplish this, putting Netscape at a disadvantage.
More directly, I'm not aware of any public Windows APIs that handed out data that tricked your code into doing a general protection fault (https://en.wikipedia.org/wiki/General_protection_fault), or in the case of WordPerfect, compel programmers to use their own printer drivers, and write code that regularly drops your tables, figures, etc. to the bottom of your document (unlike the other companies I cited, WordPerfect actually managed to ship a Windows version that didn't GP fault, but that was late, terribly buggy, and prompted all the WP fans I knew to reluctantly switch to MS Word).
People were doing Ajax-like things before XMLHTTPRequest with hidden iframes. XHR was more convenient, but it didn't bring new functionality to the table.
Netscape Navigator 4 had frustrating limitations once you had the data, but people were doing all sorts with it regardless. More was possible with Internet Explorer 4, but again, you didn't need XMLHTTPRequest. And if you just want to strip it back to the bare-bones of loading and saving data, a few people were even doing that as far back as Netscape Navigator 2 if I recall correctly – by dynamically loading new images with varying query string parameters.
One other point worth remembering: their browser, at least by version 3, was superior in speed and vastly superior in stability. They also ushered in the modern era of web applications by inventing XMLHttpRequest (https://en.wikipedia.org/wiki/XMLHttpRequest).
Netscape destroyed itself as much as Microsoft helped them by incompetence in developing software, server as well as browser. A common pattern in the '90s was Microsoft's competitors losing their ability to develop mostly working software while Microsoft didn't. That sealed the fate of Lotus, WordPerfect, Netscape, and Ashton-Tate (dBASE), and Borland did themselves no favors by buying the latter.