News The FAA seeks to eliminate floppy disk usage in air traffic control systems

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
If the app is stable running under Windows 95, a viable approach could be to run W95 and the app stack in a VM under a modern and properly secured hypervisor. The host system provides security and stability, and 95 gets all the resources it needs in a sandboxed environment. It would probably run faster than it currently does directly on old hardware. It would certainly be safer.

As dangerous as it is to run on an obsolete OS, there are risks in migration as well, especially with a service that needs near constant availability. A VM could allow this obviously reliable app to run safely without changes for years to come and to migrate seamlessly to new hardware again when the time comes.
 
Consider a scenario where you may not want to preempt your applications; one where the realtime operation of the application is more important than the OS having control over it. Now imagine that every single byte of code is perfectly tuned to the hardware it's on. In such a scenario, application crash recovery is worse than useless and a preemptive OS is less stable for the intended function of the system.
Then I might ask---why would you bother with a multitasking OS at all? Never mind one with the bloat of the 9x kernel, even the embedded one?

Wouldn't it just be simpler to use a custom built small linux or other Unix-like kernel?
 
Then I might ask---why would you bother with a multitasking OS at all? Never mind one with the bloat of the 9x kernel, even the embedded one?

Wouldn't it just be simpler to use a custom built small linux or other Unix-like kernel?
Using off the shelf will be cheaper and supporting of custom OS is very difficult. The current system has been around for 30+ years. What happens when the developers of the custom OS are all retired in 10 years and you have problems?
 
But considering MS often stops support for their OSs in the same amount of time, why would that be any better?
Outside of the change from 9X to XP causing a lot of previous software to not work due to major architecture differences in the OS most software designed on XP will work on Win 11. Upgrading the OS, while for sure a non trivial task, still allows for you to use older software. Also using commercial software makes sure that it is continually updated for security and newer technology.
 
But not the underlying OS. Which is the problem, especially for critical use things.
Since Win 7, you can easily upgrade your OS to the newer versions. For Windows Server you have been able to go to newer versions since 2016 (might have been able to do it in 2008 but I'm not positive on that). Not to mention your enterprise level OS's (Windows, SUSE, Red Hat, etc...) offer at least 8 if not 10 years of security patches for their OS. For Windows they offer 10 years even on the desktop. After 10 years you start getting to the point where the architecture of the OS is so old that patching for new technologies isn't feasible. Then you migrate to a new/er OS and you are good for a long time again. Once you get into a custom built OS you don't have the level of support, aren't guaranteed any updates, etc... As someone who has worked with custom software it is a nightmare. A custom OS would be even worse.
 
I'm not saying you'd have to build a custom OS---but you could build a small, tight linux distribution with minimalist software packages, and get the inherent advances with updates.

And we're not talking NT kernel windows (7 in your example is, of course, NT and not 9x)...that's a completely different ballgame. 9x is fundamentally flawed in the way it is built, it had to be to keep the near 100% DOS compatibility and easy migration for users from the 3.x universe. I'm not saying its Microsoft's fault for building it that way, they likely had no choice for those reasons.
 
I'm not saying you'd have to build a custom OS---but you could build a small, tight linux distribution with minimalist software packages, and get the inherent advances with updates.

And we're not talking NT kernel windows (7 in your example is, of course, NT and not 9x)...that's a completely different ballgame. 9x is fundamentally flawed in the way it is built, it had to be to keep the near 100% DOS compatibility and easy migration for users from the 3.x universe. I'm not saying its Microsoft's fault for building it that way, they likely had no choice for those reasons.
Again you are talking about a custom OS. That is just not a good idea. Use Windows Server, SUSE, Red Hat, etc... and be done with it. Don't build your own thing for it.
 
"Again you are talking about a custom OS. "

No, I'm not. What I'm talking about is a scaled down Linux distribution. In which the standard libraries would be updateable from most sources. You just don't need every single library from a general purpose distribution.

The only problem I see is that Linux was still in its fairly early lifetime during the time when the 9x kernels would have been in heavy use in embedded systems. But you can apply it to the Unixes or Unix-like that also existed already.
 
No, I'm not. What I'm talking about is a scaled down Linux distribution. In which the standard libraries would be updateable from most sources. You just don't need every single library from a general purpose distribution.

The only problem I see is that Linux was still in its fairly early lifetime during the time when the 9x kernels would have been in heavy use in embedded systems. But you can apply it to the Unixes or Unix-like that also existed already.
As soon as you start stripping things down it is a custom OS. Don't mess around with the OS at all. Just use the entire thing off the self.