Hacker News Reader

Comment by mkozlows | Hacker News Reader

The "wheel" is 40 years old. It's a great wheel, it's doing a hell of a job. But with 40 years of perspective, maybe it actually can be re-invented better without carrying around a pile of legacy back-compatibility needs.

View on HN

There are two wheels involved, the micro-kernel and capabilities, though they are roughly 40 years old.

Do you think it will ever replace Linux on something that isn't a phone?

Ars Technica got it to work on a Pixel book: https://arstechnica.com/gadgets/2018/01/googles-fuchsia-os-o...

Thanks for the link. From what we can see now it got a lot of Google integration. Too much for my taste.

Considering that they could plausibly have it shipping on tens of millions of devices as soon as they get a browser working and hardware support for a well controlled subset of modern systems, I think that's a real possibility.

If it's so terrible why did microsoft develop the windows subsystem for linux? Shouldn't they instead try to avoid the "pile of legacy" as much as possible?

Well they didn't replace their NT kernel with WSL? So it's a bit of apples-to-oranges comparison. WSL is there to provide access to the existing, vast library of linux software that might not have been treated with win32 ports?

1. MS is all about "piles of legacy." That is not counter to their philosophy at all. 2. MS apparently sees some long-term advantage to having a Linux compatibility layer. (By "advantage," I mean of course some way to make more money.)

Neither of these things has any bearing on the quality of Linux, or lack thereof.

How did you manage to read "it's terrible" into "it's a great wheel"?

Because developers who write software that is deployed to non-Microsoft cloud providers are writing code that runs on Linux, and WSL makes that a lot easier to do on Windows.

Because they saw a market opening with UNIX devs no longer happy with the hardware selection for using macOS as a pretty UNIX.

Also their goal is not to run 100% of POSIX or Linux specific software, rather achieve a good enough compatibility to run majority of well known projects and utilities.

WSL is an attempt at getting developers to use Windows rather than OS X or Linux. That's a very different situation than coming up with the fundamentals for a new operating system.

> The "wheel" is 40 years old. It's a great wheel, it's doing a hell of a job. But with 40 years of perspective, maybe it actually can be re-invented better without carrying around a pile of legacy back-compatibility needs.

Newer is rarely better. Besides, in computing, so many "new" things are just rehashes of old (and not so old) ideas.

I really wish we could get away from the "that software is old" meme in this industry.

One great advantage of old software that has been continuously developed by the same core of people is that when new features are added, they have the benefit of being informed by all the past failures and near-misses. And they are almost always features that are actually new.

New software, otoh, hasn't been through a 40-year shakedown cruise.

Old software often had requirements or made design decisions based on assumptions that aren't true anymore. These can make it more difficult to deal with or extend. I agree that in general we should just use the old software that works. However big projects like fuscia can reap some long term benefits. In the end it's a cost/benefit analysis to decide.

Well, the context that shaped old software has a habit of re-appearing, so the design decisions/trades-off may have become relevant again at least once since. The first vectorized code I saw was a relic from a long-gone Cray 1.

What? New software is usually better, or do you think people never improve anything? That just makes no sense.

Of course software improves. The context was BSD, Posix, etc are based on 40 year old tech. Those systems got a lot right and it takes a hell of a lot of hubris to think just because it’s 40 years old it is somehow not good anymore.

Reply on news.ycombinator.com