6 ms·

Google clearly has a LOT of cash to play with. They could have refactored/re-used the permissively licensed BSD core, but ended up re-inventing the wheel, break

Google clearly has a LOT of cash to play with. They could have refactored/re-used the permissively licensed BSD core, but ended up re-inventing the wheel, breaking a TON of POSIX based software. For what benefit exactly?

tpush 9y ago
??? The whole point of this OS are concepts like capabilities that aren't represented in POSIX interfaces.
tmikaeld 9y ago
Complete and total lock-in?
- jasonvorhe 9y ago
  That must be why the code is on GitHub.
  - martin1975 9y ago
    So is Linux' code, except if you try to redistribute it by linking to it, you'd have to disclose your mods. Not so with Fuchsia's license. They can push out proprietary code all they want.
- martin1975 9y ago
  They certainly picked the correct license to do just that. GenodeOS has exactly what they needed on a formally proven kernel. Why didn't they use that? NIH syndrome... that's why. And of course, control.
lainga 9y ago
Promotions at Google are based on quality of implementation, not original ideas.
- forgottenpass 9y ago
  The history of chat products out of google shows that it's about quantity, not quality.
  - calebh 9y ago
    Yup, I'm wondering how many more days Allo has to live. I don't even bother with the proprietary Google messenger apps anymore because I know they'll be killed within a couple years or less.
butterisgood 9y ago
I'm sure people said that about Bell Labs and Unix too.
mtgx 9y ago
I read in an older article that POSIX software compatibility may infringe on Oracle's copyrights. Maybe that's why?
- AdmiralAsshat 9y ago
  Making a piece of software that follows the POSIX standard, a term that Richard Stallman came up with, runs afoul of copyright? That would seem like a massive slap in the face to the cause of freedom, if so.
  - sp332 9y ago
    It would be ironic. But RMS only proposed the name, not the standard itself.
izacus 9y ago
There's nothing implying that Google wants to make a POSIX compliant OS. Why do you think that's important for a Fuchsia type project? The other huge Google OS (Android) doesn't benefit from POSIX either at all - the POSIX interface isn't even exposed to the user applications.
- saagarjha 9y ago
  Well, not the full POSIX interface. Bionic does expose a significant portion, though.
  - pjmlp 9y ago
    It is not part of the official NDK APIs, use it at your own peril, regarding NDK lockdown. ANSI C and POSIX do overlap, but aren't the same thing.
  - izacus 9y ago
    In NDK yes. Those aren't really as fully compliant as we'd wished (and differences in implementation are sometimes really annoying to deal with). Nevertheless, vast majority of applications will never include NDK code and even if they will, they probably won't really talk to POSIX calls directly either. For the type of OS Android is, it doesn't really gain anything by being POSIX compliant far away from developer APIs. And Fuchsia, if anything, is trying to be an Android type OS.
- mankash666 9y ago
  Well - requiring basic frameworks to be re-written for your vanity-OS is the fastest path to irrelevance. Most libraries and frameworks (from curl to zlib/git) assume a POSIX interface. If Fuchsia doesn't care about these fundamental utilities being functional, then that's a flawed assumption regarding the developer ecosystem
  - saagarjha 9y ago
    Assuming Google is even targeting the developer ecosystem here. I'd assume that most of the users of Fuchsia will be end users, so that's what Google is targeting.
forgotmypw 9y ago
For the benefit of one or a few individuals "burning through a lot of cash", "spearheading the effort", and "padding their resume(s)".
- mankash666 9y ago
  I don't know why they down-voted you. The pitch, if any, for fuchia, wouldn't stand muster outside Google. Clearly, some folks benefit from "burning through a lot of cash", "spearheading the effort", and "padding their resume(s)"
  - tomc1985 9y ago
    Why not? Many are tired of Linux's constant convention thrashing. I would love to reimagine Linux minus all its ancient, historical cruft. Yes there's BSD, but it is nice to have another take.
    - mankash666 9y ago
      I don't know if you see the irony in your arguments. You demonize capitalism in another comment to me, and argue here that another take that requires a LOT of funding is a "nice to have". Do you realize that this "nice to have" is made possible by the deep coffers of Google, whose existence is owed to capitalism?
      - majewsky 9y ago
        Non sequitur. Capitalism is not the only (possible/existing) deep-coffer generator. See China.
      - tomc1985 9y ago
        Yeah I'm not really agreeing with ya there. Whether its put together by Google, or the community, or whoever, really doesn't matter. Yes in this case it's Google, who does exist due to capitalism (not exclusively due to capitalism, though, as other systems work in other countries), but then capitalism is not the only economic or political system that produces 'deep coffers'. And this 'nice to have' is not build-able exclusively by Google. By your logic anyone born to a family where the parents have U.S. employment would owe their existence to capitalism. Does that instinctively seem correct?
amaccuish 9y ago
Umm, with android apps being written in java, I don't think most developers or users worry about POSIX, and Linux isn't fully posix anyway.
- saagarjha 9y ago
  > with android apps being written in java Many Android apps drop down to the NDK, where POSIX does matter. > Linux isn't fully posix anyway It's close enough that the difference really is just an issue of not getting certification.
  - pjmlp 9y ago
    From POSIX, only the thread APIs are officially supported on Android. Everything else on the NDK has nothing to do with POSIX. https://developer.android.com/ndk/guides/stable_apis.html https://developer.android.com/ndk/guides/stable_apis.html
- tracker1 9y ago
  Not to mention other higher level languages that have an abstract for io. (libuv in node and elsewhere for example)
wbl 9y ago
Better security model when number of users per system is less than unity.
mkozlows 9y ago
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.
- JustSomeNobody 9y ago
  > 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.
  - jay-anderson 9y ago
    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.
    - gnufx 9y ago
      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.
  - IshKebab 9y ago
    What? New software is usually better, or do you think people never improve anything? That just makes no sense.
    - JustSomeNobody 9y ago
      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.
willvarfar 9y ago
Google invest heavily in securing computing generally, from creating the secure browser Chrome because the alternatives at the time were very insecure to a big paid team of security researchers (recent successes include discovering Spectre and Meltdown) to creating secure operating systems to possibly eventually replace Android and ChromeOS. A lot of security-interested programmers think it easier to secure a small microkernel like Fuchsia than to secure the older generation of operating systems and the baggage they have acquired. Google want everyone to trust the Internet. If consumers are scared to shop online, Google dies. This is why they put so much effort into securing it.
globuous 9y ago
Would it be possible to conceive a POSIX layer of compatibility for Fuchsia?
- JdeBP 9y ago
  Yes. https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742
jjrh 9y ago
> For what benefit exactly? Wouldn't surprise me if the unofficial reason is because it's more fun.
dreta 9y ago
Personally, i'll be happy if we can move past read/write constructs designed for tape drives.
- kps 9y ago
  Networks are the new tapes — still slow and sequential.
noselasd 9y ago
Why would it break a ton of posix software? They seem to have ported the musl c library, they removed to have stuff like cmake vim git openssh running on it

tpush 9y ago

??? The whole point of this OS are concepts like capabilities that aren't represented in POSIX interfaces.

tmikaeld 9y ago

Complete and total lock-in?

jasonvorhe 9y ago
That must be why the code is on GitHub.
- martin1975 9y ago
  So is Linux' code, except if you try to redistribute it by linking to it, you'd have to disclose your mods. Not so with Fuchsia's license. They can push out proprietary code all they want.
martin1975 9y ago
They certainly picked the correct license to do just that. GenodeOS has exactly what they needed on a formally proven kernel. Why didn't they use that? NIH syndrome... that's why. And of course, control.

lainga 9y ago

Promotions at Google are based on quality of implementation, not original ideas.

forgottenpass 9y ago
The history of chat products out of google shows that it's about quantity, not quality.
- calebh 9y ago
  Yup, I'm wondering how many more days Allo has to live. I don't even bother with the proprietary Google messenger apps anymore because I know they'll be killed within a couple years or less.

butterisgood 9y ago

I'm sure people said that about Bell Labs and Unix too.

mtgx 9y ago

I read in an older article that POSIX software compatibility may infringe on Oracle's copyrights. Maybe that's why?

AdmiralAsshat 9y ago
Making a piece of software that follows the POSIX standard, a term that Richard Stallman came up with, runs afoul of copyright? That would seem like a massive slap in the face to the cause of freedom, if so.
- sp332 9y ago
  It would be ironic. But RMS only proposed the name, not the standard itself.

izacus 9y ago

There's nothing implying that Google wants to make a POSIX compliant OS. Why do you think that's important for a Fuchsia type project? The other huge Google OS (Android) doesn't benefit from POSIX either at all - the POSIX interface isn't even exposed to the user applications.

saagarjha 9y ago
Well, not the full POSIX interface. Bionic does expose a significant portion, though.
- pjmlp 9y ago
  It is not part of the official NDK APIs, use it at your own peril, regarding NDK lockdown. ANSI C and POSIX do overlap, but aren't the same thing.
- izacus 9y ago
  In NDK yes. Those aren't really as fully compliant as we'd wished (and differences in implementation are sometimes really annoying to deal with). Nevertheless, vast majority of applications will never include NDK code and even if they will, they probably won't really talk to POSIX calls directly either. For the type of OS Android is, it doesn't really gain anything by being POSIX compliant far away from developer APIs. And Fuchsia, if anything, is trying to be an Android type OS.
mankash666 9y ago
Well - requiring basic frameworks to be re-written for your vanity-OS is the fastest path to irrelevance. Most libraries and frameworks (from curl to zlib/git) assume a POSIX interface. If Fuchsia doesn't care about these fundamental utilities being functional, then that's a flawed assumption regarding the developer ecosystem
- saagarjha 9y ago
  Assuming Google is even targeting the developer ecosystem here. I'd assume that most of the users of Fuchsia will be end users, so that's what Google is targeting.

forgotmypw 9y ago

For the benefit of one or a few individuals "burning through a lot of cash", "spearheading the effort", and "padding their resume(s)".

mankash666 9y ago
I don't know why they down-voted you. The pitch, if any, for fuchia, wouldn't stand muster outside Google. Clearly, some folks benefit from "burning through a lot of cash", "spearheading the effort", and "padding their resume(s)"
- tomc1985 9y ago
  Why not? Many are tired of Linux's constant convention thrashing. I would love to reimagine Linux minus all its ancient, historical cruft. Yes there's BSD, but it is nice to have another take.
  - mankash666 9y ago
    I don't know if you see the irony in your arguments. You demonize capitalism in another comment to me, and argue here that another take that requires a LOT of funding is a "nice to have". Do you realize that this "nice to have" is made possible by the deep coffers of Google, whose existence is owed to capitalism?
    - majewsky 9y ago
      Non sequitur. Capitalism is not the only (possible/existing) deep-coffer generator. See China.
    - tomc1985 9y ago
      Yeah I'm not really agreeing with ya there. Whether its put together by Google, or the community, or whoever, really doesn't matter. Yes in this case it's Google, who does exist due to capitalism (not exclusively due to capitalism, though, as other systems work in other countries), but then capitalism is not the only economic or political system that produces 'deep coffers'. And this 'nice to have' is not build-able exclusively by Google. By your logic anyone born to a family where the parents have U.S. employment would owe their existence to capitalism. Does that instinctively seem correct?

amaccuish 9y ago

Umm, with android apps being written in java, I don't think most developers or users worry about POSIX, and Linux isn't fully posix anyway.

saagarjha 9y ago
> with android apps being written in java Many Android apps drop down to the NDK, where POSIX does matter. > Linux isn't fully posix anyway It's close enough that the difference really is just an issue of not getting certification.
- pjmlp 9y ago
  From POSIX, only the thread APIs are officially supported on Android. Everything else on the NDK has nothing to do with POSIX. https://developer.android.com/ndk/guides/stable_apis.html https://developer.android.com/ndk/guides/stable_apis.html
tracker1 9y ago
Not to mention other higher level languages that have an abstract for io. (libuv in node and elsewhere for example)

wbl 9y ago

Better security model when number of users per system is less than unity.

mkozlows 9y ago

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.

JustSomeNobody 9y ago
> 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.
- jay-anderson 9y ago
  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.
  - gnufx 9y ago
    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.
- IshKebab 9y ago
  What? New software is usually better, or do you think people never improve anything? That just makes no sense.
  - JustSomeNobody 9y ago
    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.

willvarfar 9y ago

Google invest heavily in securing computing generally, from creating the secure browser Chrome because the alternatives at the time were very insecure to a big paid team of security researchers (recent successes include discovering Spectre and Meltdown) to creating secure operating systems to possibly eventually replace Android and ChromeOS. A lot of security-interested programmers think it easier to secure a small microkernel like Fuchsia than to secure the older generation of operating systems and the baggage they have acquired. Google want everyone to trust the Internet. If consumers are scared to shop online, Google dies. This is why they put so much effort into securing it.

globuous 9y ago

Would it be possible to conceive a POSIX layer of compatibility for Fuchsia?

JdeBP 9y ago
Yes. https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742 https://news.ycombinator.com/item?id=16816742

jjrh 9y ago

> For what benefit exactly? Wouldn't surprise me if the unofficial reason is because it's more fun.

dreta 9y ago

Personally, i'll be happy if we can move past read/write constructs designed for tape drives.

kps 9y ago
Networks are the new tapes — still slow and sequential.

noselasd 9y ago

Why would it break a ton of posix software? They seem to have ported the musl c library, they removed to have stuff like cmake vim git openssh running on it