3 ms·

> so actual Flutter is just Google NIH syndrome… So true. If Google can't control it, they build a competitor. If Google can't figure it out, or if it is not

> so actual Flutter is just Google NIH syndrome… So true. If Google can't control it, they build a competitor. If Google can't figure it out, or if it is not 100% what they need, they will build a competitor. This has caused tons of problems for both the industry and consumers. For example, Android updates would be much less of a mess if Google had used linus' LTS kernel...

vlovich123 6y ago
How does sticking to an LTS solve the lack of stable driver ABI? That’s what project Treble from Google was about - providing a stable ABI driver authors/vendors can target so that an OS update can build and function correctly. Also I don’t know how closely you followed the ARM tree of the Linux kerne but pre-Google the authors were resistant to the board config approach because you don’t have that on x86 because there isn’t an SoC that bundles a lot of the functionality. It wasn’t until Android that the Linux tree started taking ARM more seriously (which arguably it wasn’t really before anyway)
- panpanna 6y ago
  Well, for one thing you have only one ABI to consider if you go with the LTS instead of hundreds we have today. Treble is Google's attempt to fix a problem they - at least mostly - created themselves thanks to their NIH attitude. Not sure I would attribute use of device trees to Google. I think the ARM-fication of the kernel happened a few years before Android.
  - vlovich123 6y ago
    You may want to brush up on your history: https://elinux.org/Device_tree_history https://elinux.org/Device_tree_history Russell King as recently as 2009 was against it. It got merged around ~2013 so it took about 4 years. The first Android phone was supposed to be 2009 but IIRC it got delayed to 2010 because of the iPhone (announced 2008). I think my characterization of this is fair and accurate. > Well, for one thing you have only one ABI to consider if you go with the LTS instead of hundreds we have today. I don't follow. You have 1 Android release per year. There is a new LTS kernel every year. I don't see how sticking to LTS solves the problem. You have to actively hold off on Linux kernel releases to get any advantage of this process. For something like Android this means they'd be forking LTS & it would be even more work to move to a new kernel (all the out-of-tree development + all your driver vendors switching kernels). So now you'd be complaining why Android is using such an ancient Android kernel. Treble solves a real problem and has nothing to do with NIH unless you just define all kernel development happening at Google as NIH (which would be strange considering how many Linux kernel developers and maintainers they employ).

vlovich123 6y ago

How does sticking to an LTS solve the lack of stable driver ABI? That’s what project Treble from Google was about - providing a stable ABI driver authors/vendors can target so that an OS update can build and function correctly. Also I don’t know how closely you followed the ARM tree of the Linux kerne but pre-Google the authors were resistant to the board config approach because you don’t have that on x86 because there isn’t an SoC that bundles a lot of the functionality. It wasn’t until Android that the Linux tree started taking ARM more seriously (which arguably it wasn’t really before anyway)

panpanna 6y ago
Well, for one thing you have only one ABI to consider if you go with the LTS instead of hundreds we have today. Treble is Google's attempt to fix a problem they - at least mostly - created themselves thanks to their NIH attitude. Not sure I would attribute use of device trees to Google. I think the ARM-fication of the kernel happened a few years before Android.
- vlovich123 6y ago
  You may want to brush up on your history: https://elinux.org/Device_tree_history https://elinux.org/Device_tree_history Russell King as recently as 2009 was against it. It got merged around ~2013 so it took about 4 years. The first Android phone was supposed to be 2009 but IIRC it got delayed to 2010 because of the iPhone (announced 2008). I think my characterization of this is fair and accurate. > Well, for one thing you have only one ABI to consider if you go with the LTS instead of hundreds we have today. I don't follow. You have 1 Android release per year. There is a new LTS kernel every year. I don't see how sticking to LTS solves the problem. You have to actively hold off on Linux kernel releases to get any advantage of this process. For something like Android this means they'd be forking LTS & it would be even more work to move to a new kernel (all the out-of-tree development + all your driver vendors switching kernels). So now you'd be complaining why Android is using such an ancient Android kernel. Treble solves a real problem and has nothing to do with NIH unless you just define all kernel development happening at Google as NIH (which would be strange considering how many Linux kernel developers and maintainers they employ).