Ubuntu 9.10 Off To A Great Performance Start

[suokko]

'Stable APIs' are a really, really bad idea from a technical standpoint, because it ties you into bad design choices made years before. I don't know why this is so hard for some people to understand.

It is simple to change underlying implementation while keeping same interface. That why we have libraries between kernel and programs translating the SUS interface to kernel interface.

[movieman]

It is simple to change underlying implementation while keeping same interface.

You can do it, yes: Windows ability to run ancient DOS programs has proven that. But why would you want to?

Again, the only reason to push a 'stable API' on the kernel is because you don't want to release your source code. If you don't want to release your code, then the onus is on _YOU_ to keep up with kernel changes, not on kernel developers to go to a great deal of extra trouble maintaining APIs that they no longer want.

[deanjo]

You can do it, yes: Windows ability to run ancient DOS programs has proven that. But why would you want to?

Why? Because there are a large number of applications out there that are specialized for tasks. DOS based apps are still alive and well in many industrial applications. Typically these applications are not your "on the shelf for $50" but often cost thousands of dollars and do their job rather well. Just off the top of my head I can think of 10+ applications that there is no other OS equivalent. Hardware gets old and dies and replacing it shouldn't mean having to re-invest in software that fully accomplishes the task needed.

[kraftman]

Why? Because there are a large number of applications out there that are specialized for tasks. DOS based apps are still alive and well in many industrial applications. Typically these applications are not your "on the shelf for $50" but often cost thousands of dollars and do their job rather well. Just off the top of my head I can think of 10+ applications that there is no other OS equivalent. Hardware gets old and dies and replacing it shouldn't mean having to re-invest in software that fully accomplishes the task needed.

There wouldn't be such problems with open source. It's better to implement better API from time to time and OS projects will adapt. You don't have to keep crappy designed ABI's to make some old programs work. It's one of the reasons why OS is better then closed source.

[deanjo]

There wouldn't be such problems with open source. It's better to implement better API from time to time and OS projects will adapt. You don't have to keep crappy designed ABI's to make some old programs work. It's one of the reasons why OS is better then closed source.

Not true at all. How many developers for example would maintain items such as dry cleaning software, oil field monitoring software, trucking software, ILS systems, weight and scale systems, custom inventory applications, seed cleaning software, etc etc etc. Many such applications would require complete rewrites add large cost to the user for a system that would mean either investing heavily in not only paying for the redevelopment of the software but also in many industrial cases would require upgrading of the hardware they are meant to be monitoring as well. When it comes to opensource such solutions are like trying to find hens teeth where they are maintained on a regular basis. Being opensource by no means of the imagination means maintained. Sure one might have the oppertunity to do so but it does not mean by any extent that will happen, especially in specialized software where the target market is small.

[kraftman]

Not true at all. How many developers for example would maintain items such as dry cleaning software, oil field monitoring software, trucking software, ILS systems, weight and scale systems, custom inventory applications, seed cleaning software, etc etc etc. Many such applications would require complete rewrites add large cost to the user for a system that would mean either investing heavily in not only paying for the redevelopment of the software but also in many industrial cases would require upgrading of the hardware they are meant to be monitoring as well. When it comes to opensource such solutions are like trying to find hens teeth where they are maintained on a regular basis.

If they have to be completely rewritten that's clear. If such apps are very old it will be even harder to maintain them or impossible if they're closed source and company is no more.

[deanjo]

If they have to be completely rewritten that's clear. If such apps are old it will be even harder to maintain them or impossible if they're closed source and company is no more.

Thus why maintaining that backward capability is important. To switch to other solutions would millions if not billions of dollars in infrastructure to redo. Saying it sucks because it's closed and hoping that someone may even have the faintest clue let alone ambition to maintain such needs exists in a "dream world" but when real life is in play it doesn't apply in practice.

[kraftman]

Thus why maintaining that backward capability is important. To switch to other solutions would millions if not billions of dollars in infrastructure to redo.

When comes to unstable API/ABI you can usually stick with older system which remains compatible with some application.

Saying it sucks because it's closed and hoping that someone may even have the faintest clue let alone ambition to maintain such needs exists in a "dream world" but when real life is in play it doesn't apply in practice.

Yeah, theory sometimes looks nice, but it's not so nice when comes to real situation. However, you still have greater chance to do something which such old relic if it's open source. Everything depends

[susikala]

Thus why maintaining that backward capability is important. To switch to other solutions would millions if not billions of dollars in infrastructure to redo. Saying it sucks because it's closed and hoping that someone may even have the faintest clue let alone ambition to maintain such needs exists in a "dream world" but when real life is in play it doesn't apply in practice.

Your point is largely moot. Saying that crappy approaches to software design should be kept in place and used because they have met with large audience and _now_ the only affordable solution is to keep them in place is no more than saying 'this is the situation, face it'. It's not even an argument, it's at best a laying forth of the pitiful situation in the closed source world.

The 'solution' to this 'situation' is developing sustainably to begin with, and that would mean for Linux.

[deanjo]

Your point is largely moot. Saying that crappy approaches to software design should be kept in place and used because they have met with large audience and _now_ the only affordable solution is to keep them in place is no more than saying 'this is the situation, face it'. It's not even an argument, it's at best a laying forth of the pitiful situation in the closed source world.

The 'solution' to this 'situation' is developing sustainably to begin with, and that would mean for Linux.

Perhaps we should apply the same philosophy to hardware in the kernel too then. Think of all the additional talent that could be used in focusing on current hardware instead of maintaining old legacy hardware. Think of how much easier it would be to maintain the kernel if all that old hardware support was scrapped.