- mrpijey: Except you hit the problem right there - virtualization relies on the host hardware. So the moment Intel makes some breaking change somewhere, Windows 95 may no longer run under a virtualizer. And it also means it is beyond the hypervisor developer's control. Hence why hypervisors are mainly used to run OS'es from the same period as the host hardware. Sure, there were some that could also run older software, such as Virtual PC 2007, but that came as a bonus rather than an intended feature, and was something was being actively phased out, eg. Virtual PC 2007 dropped official support for MS-DOS, and Windows Virtual PC 2007 dropped official support for anything earlier than Windows XP. For running anything older, an emulator becomes the requirement.
And no, emulators won't die. For exame, MAME was started over 15 years ago and it's still actively developed. And with most present-day emulators being open-source, even when the active developers go away, there will always be someone to take over. So if, say, by 2025, Tom Walker and I had both lost interest in emulation and the Microsoft releases, say, Windows 14 which breaks PCem and PCem-X, someone can take the source code and make the emulator work again. Same is also true for some hypervisors, such as VirtualBox, which happen to be open-source. But for example, Virtual PC no longer works on Windows 10, and since only Microsoft have the sources and have decided to not fix it, the product dies there. And if VMWare eventually folds and subsequently a Windows version breaks VMWare, well, the product dies there too.
So both emulators and hypervisors have an equal likelihood of dying or surviving.
The more advanced systems become the more difficult they will become emulated, while virtualizing it is a fairly simple matter with minimal overhead to the performance. So no, hypervisors won't go away for some time, and emulators will die far long before hypervisors will. It's still a very complicated to emulate a PowerPC system on a PC today (extremely slow and buggy), heck, even emulating an old Nintendo 64 is a buggy hit and miss today, while every modern computer today can easily hypervise any x86 OS you want, with minimal speed penalty.
Except, the difference is, emulators are designed to provide us virtual versions of systems from the past, while hypervisors are designed to provide us virtual versions of systems of the present. The usage cases are completely different.
Also, emulation will evolve too. Some day, someone will come up with better emulation code. Host hardware will also evolve, so emulators that are now slow will then become faster. And so on. There's also the fact, as you rightfully point out, that hypervisors tend to be developed by corporations which can invest in full-time development of them, while most emulators tend to be hobby projects. But that's not the developers' fault now, is it? How many people have you seen willing to make even a small monetary donation to an emulator developer? Yep, that's right, none. So yeah, if we have to spend time looking for another job to make a living, you can't expect us to progress as fast as corporate-developed hypervisors progress.
But as I said, emulation will keep improving. And as I said, the usage cases of emulators and hypervisors are completely different. The former are used principally for either entertainment or software archaeology, while the latter are used principally for running modern software. And here we are talking about software archaeology, ie. running old software for the purpose of learning from it. For that, emulators win hands down as they provide a virtual version of the hardware the software we are talking about is expecting to be run on, even if said virtual version is imperfect. Hypervisors were only good for software archaeology while host hardware was still reasonably close to period hardware, but the gap is increasing, and I gave a whole list of what doesn't run on Virtual PC 2007 for example, and the list gets longer for VMWare and is the longest for VirtualBox.
Edit: And I just remembered, that even in the usage cases hypervisors are primarily designed for, emulation is gaining a foothold. Specifically, QEMU (or KVM which is based on QEMU) seems to be widely used in all sorts of settings, and it seems it works just as well as hypervisors, if not better because it's multiplatform (ie. can just as easily run on ARM, IA-64 (Itanium), etc.), and provides much more predictable behavior of guest OS'es. Of course, it performs better than PCem-X for example, but it's designed for a vastly different usage case. It is still however an emulator.