Computer Backup

[Paul]

"Full-system emulation. Run operating systems for any machine, on any supported architecture."

There is some sort of USB passthru. It might not be available on every platform (the passthru might be a separate Github project).

But if you get that actually working for a customer, that's Nobel Prize material :-)

While a lot of the easy-peasy Virtual Machine hosting is homogeneous (like x86-on-x86), QEMU should be able to do heterogeneous cases (arm-on-x86, x86-on-arm).

When we used Softwindows at work, that was x86-on-Sparc. And it was slow, because... the CPU only ran at 400MHz or so. Some of the staff got "PC-on-card" hardware, and their Windows capabilities on Sparc, ran at something more like a normal speed.

I think we used a floating license for Softwindows at the time, because normally people didn't run PC software all that often. You might get some advertising material (.doc) from a local salesman to print off, and that might take a few minutes of a license.

Performance:

Hobby level, brute force, heterogeneous 0.01x

Code loop translated, cached, heterogeneous 0.1x

Virtualbox (homogeneous) before they broke it 0.9x [5GHz processor feels like 4.5GHz]

This leaves quite a range of performances, and typically is why things like x86-on-arm are wheezing on you. There is more than one reason you don't want to do that, without training wheels on :-)

When I ran QEMU KVM on Linux, which is an "easy case", it took me all day, looking at various conflicting documentation, as to which CPU choice I should be using for "transparent" translation (as close to x86-on-x86 as you could get). There is a menu with a hundred items in it, and only two choices make any sense. And one of the choices is supposed to crash and stuff. Well, it didn't crash actually. So that is why it is Nobel Prize material, you get it working in spite of the docs.

Paul

[Paul]

If an older scanner cost enough money (one of the high res slide scanners), then something like Vuescan from

might have worked.

If the scanner costs as much as Vuescan does, then it's hardly worth trying to solve it that way. Some cheap new scanners, are very close in price to Vuescan.

Still, for cases where you think your old scanner cannot possibly work on a new OS -- try it! You might surprise/shock yourself.

For some of the scanners, you can check for family inheritance, to see if possibly a friends model which is slightly different than yours, actually has the same hardware. I like to check items like LiDE 50 on here, to see which models were the same.

Paul

[wasbit]

I agree, but the posters don't ask for, or give, a solution & so there post is basically a troll.

[The Natural Philosopher]

SANE is what has been recompiled by *someone else*.

[John Rumm]

Yeah I did look at that, and decided I did not want to dive into that rabbit hole at the time :-)

I think some of the paid for options (parallels et al) might support cross architecture on arm, but when you suggest they might want to pay for a copy, they are less keen!

As an aside, usbip looks interesting:

I have seen it working well with linux at both ends, but there is a windows port as well now. Presumably it will work in a mixed platform environment as well. Could be quite nice for things like USB document scanner connected to a Raspberry Pi, with the USB exported and then mounted on a windows box via the LAN.

Sometimes it can be interesting to see how many emulation layers deep of you can go. Back in the late 80's (long before things like hypervisors, and nested virtualisation) SWMBO and I wrote a 6502 emulator for our final year project at university. It was originally written in Pascal on their Prime mini computer. Later I recompiled it in Turbo Pascal for the PC and it ran in DOS. Then I ran a software[1] PC emulator on my Amiga

2000, and ran my emulator in that.

So a ~7MHz 68K emulating x86, emulating 6502. It worked, but the execution speed was certainly not measured in MIPS!

[1] There were also hardware PC "bridge boards" (basically a PC on a card) which would drop into one of the Zorro slots that was in line with one of the ISA bus slots that it also had on the motherboard.

[John Rumm]

Don't know if you watch Dave's Garage on youtube, but he (Dave Plumber - original author of task manager[1]) has been walking through the source code for the XP version...

[1] A home project that he built for his own use, but enough people found it handy and eventually they asked if he would be prepared to contribute it to the windows source tree.

[John Rumm]

I got caught like that with my rather tasty Epson GT-8500 SCSI scanner, going from Win 95/98 to Win 2K (i.e. jumping architecture from DOS subsystem to the NT one). No official drivers. There was a 3rd party company that was selling them - but at more than the price of a new scanner!

[John Rumm]

I always liked the (old) windows scanner driver "TWAIN" acronym... Toolkit Without An Interesting Name (plus all that talk of TWAIN drivers made it sound like loco drivers with a lisp!)

[Andy Burns]

Here, Process Explorer rarely shows CPU > 1% at idle, Task Manager rarely shows CPU < 6% at idle

[Andrew]

Apart from 486 processors

[wasbit]

Edit ; their post

[The Natural Philosopher]

I first saw Linux on a 486.

Support only ceased a year ago. There are old distro versions that will run.

And in any case that is a straw man. I was making the point that a modern 64bit linux can handle older peripherals with ease. Not that a modern 64 bit OS could be made to run on a 34 years old 32 bit chip.

[Andrew Gabriel]

I just had a problem with firefox corrupting its profile directory somewhere such that it wouldn't run properly. I just rolled it back to yesterday's snapshot, and it's working perfectly again.

[John Rumm]

You can do similar things on NTFS using the volume shadows copy service... creating automatic roll back points.

The problem with relying on it is that the crypto scammers have learnt to look for and knobble the service and shadow copies before doing their stuff.