Hacker News Reader

Comment by geocrasher | Hacker News Reader

If MS were to write a new bespoke OS that requires new everything and maintains no backward compatibility (that's the point of a new OS, isn't it?) and launch it to the world, the amount of people who'd rush out and buy it can be counted on one hand. It could be absolutely revolutionary, but it would lack one major thing:

Momentum.

Look at every OS that's on the market right now. It solves a problem. Right now there are very few people who can't pick an OS off the shelf that will solve their particular problem. Be it an RTOS or a fancy gui that your grandma is comfortable with, the software already exists.

I guess what I'm really trying to say is that unless you can upend the current OS landscape while maintaining compatibilty, then a new OS for the masses doesn't solve a problem. It creates one.

View on HN

Though in name it's not a new OS, Windows 8, 10 are new experiences for anyone who uses them since Windows 7.

I think you're right about momentum, MS isn't able to come out of it's momentum of the old into the new.

For example, Apple transitions to a new OS with confidence and takes the ecosystem with it in a blink of an eye because they are trying control their momentum instead of being driven by it.

Edit: Windows can maintain backward compatibility with its ecosystem if they did it well, like Apple has done.

> Edit: Windows can maintain backward compatibility with its ecosystem if they did it well, like Apple has done.

Backwards compatibility with the ecosystem, especially the enormous ecosystem of software that doesn't have active maintainers chomping at the bit to test/patch/update with every single OS change, is a HUGE part of why people use Windows.

Apple really doesn't have this. Instead, they have a critical mass of software that is inside their RDF and actively updates to every breaking change they roll out. They have the "luxury" of not having to care if they break random bits of old software* in the process.

(* software that may be niche, may not have large individual user bases, but which might actually not have modern up-to-date drop-in replacements)

That's fair. I don't mean that Windows should be replaced. I mean that Windows should do what Windows is good at.

Trying to introduce Linux into Windows isn't one of these things. Doing a UI refresh is fine but trying to make Windows UI conform to spatial computing isn't going to help it or the ecosystem in the long term.

If you look at their app building documentation, it's extremely disconcerting. This much compatibility overhead isn't helping the ecosystem grow.

BSD, IllumninOS, IBM z/OS did it first in supporting Linux APIs in some form, either by syscall remapping or VMs.

Linux compatibility has become more relevant than POSIX, and the easiest way to achieve it is just to bundle it for the ride.

In Microsoft's case, they want to cater to the market that buys Apple devices, to actually develop GNU/Linux software and couldn't care less about Apple's ecosystem, aren't happy with it, and don't bother to support Linux OEMs.

It seems to be working.

> BSD, IllumninOS, IBM z/OS did it first in supporting Linux APIs in some form, either by syscall remapping or VMs.

z/OS has only gained support for running Linux binaries quite recently, in z/OS 2.4 (released September 2019), which supports running z/Linux Docker containers (zCX). Prior to that, z/OS had no built-in support for running Linux binaries in any form. Microsoft released WSL1 in August 2016. So z/OS got this feature 3 years after Windows did.

You might be thinking of z/VM or PR/SM, both of which support running z/Linux virtual machines (but neither of which is z/OS); or of z/OS Unix System Services, which offers some degree of source compatibility with Linux (through its implementation of the UNIX 95 standard), but doesn't have any Linux binary compatibility.

Yeah I guess those ones, as I remember IBM mainframes have had support for Aix/Linux guests for a while now.

Thanks for the correction.

> as I remember IBM mainframes have had support for Aix/Linux guests for a while now

IBM VM has supported Linux guests ever since Linux was first ported to IBM mainframes, circa 1999-2000. Indeed, it is quite likely the initial port was done using VM.

IBM offered two versions of AIX on mainframes, AIX/370 in 1998 and AIX/ESA in 1991. However, how much code did these have in common with AIX on RS/6000? In the later case, very little – AIX/ESA is actually a port of the Mach kernel based OSF/1 to IBM mainframes, with little or no code shared with RS/6000 AIX (which contemporary AIX descends from). It actually had more in common with Digital/Compaq/HP's discontinued Tru64 UNIX (which is another OSF/1 derivative) than with AIX on RT-PC / POWER. I get the impression that AIX/370 had little or no code in common with RS/6000 AIX either – it was developed by Locus Computer Corporation – but I don't know enough about it to be sure about that. Neither AIX/370 or AIX/ESA ever saw any great adoption – both were discontinued in the 1990s, they were replaced by MVS OpenEdition (later OS/390 OpenEdition, and now z/OS Unix System Services) and also later on also by Linux.

Thanks once more. :)

> AIX/370 in 1998

Typo: 1989 not 1998

Reply on news.ycombinator.com