AM (myname4rwt.delete@this.jee-male.com) on August 18, 2018 12:49 pm wrote:
>
> I agree with many of the points. Furthermore, there was a big ARM-related thread here a month
> or two ago, and I intensely argued for the need for bolder moves on behalf of ARM folks, mainly
> phonemakers as they are in the best position to bring always-connnected thin & light gadgets
> with full-day battery life, into larger territories, ultrabooks in the first place. So we don't
> disagree here, and I don't disagree with Linus in this regard, my reply to his post was spurred
> by exactly what I said -- I was pretty surprized to see him argue in the way he did.

AM,

I got from your original comments, that there was a lot of good points in there that you made. And I know that the odd blow-out, is the kind of thing that is needed occasionally between very talented engineers. I tend to stand back and let them at it, when it happens. Engineering is one of those things, that is an umbrella with a lot of different followers underneath it. The kernel developers floating just above the level of hardware, but in reality needing to delve right down into it. Even more so in the Linux community, because there are so many different hardware platforms involved at any one time. Which made me even more surprised that the Linux people are still waiting for these ultrabook type devices. I'm amazed.

I think the way that it works, is that some individual who might be working with their ultrabook, and intensively focused on some very specific aspect of kernel development, application development or application maintenance, will find something that makes them do a 'double take'. Like, am I really seeing what I'm seeing kind of thing. What needs to happen at that point, is one needs to take it to someone (it could be Linus, it could be dozens of others), who are seeing more of the entire picture. Ask them, can you fire up your ultrabook piece of hardware on your end and check this out for me too? I.e. This chap who in ordinary, everyday course of events would have no call to contact the 'higher up's', but because the ARM tech is now baked into something like a chip that runs on an ultrabook they are both able to work on top of something that establishes a very common baseline.

(Except as I said, one person is dealing with application maintenance perhaps, and often the closer you look, the less you see, kind of problem. The other, has an overview, but it dependent on others to help focus in on details)

So, one really does need a proliferation of some kind of 'hardware' out there, that all of these people at different levels, can use to establish a baseline relationship, between each other's projects, and each other's focus and dedication. Hardware actually is good at establishing that baseline between such a distributed collaboration effort. Because Linus might happen to be running 'X' distribution of his OS, with such as such kernel revision. The guy doing the application maintenance might be running something else. The baseline that connects them together therefore is, 'the hardware'.

And it's important too, that they're all running slightly different flavors. Because of the bug, is that important, chances are it'll show itself in a lot of different blends of the operating system. The chances are too, I'd imagine, is that it will be some obscure guy out in the forest of developers working, doing something like application maintenance that's sees something before any one else does. Because he or she, is literally walking back and forth on the same pathway, day in and day out. So anything 'out of the ordinary' just jumps out at them, at a lot faster.

It's the same point that Linus made for example about memory and very basic levels of error correction. Ideally, it would be an ultrabook with an ARM process and basic ECC memory technology built in. Not because memory errors will be very common. But because one can discount that possibility without undue delay, and focus in on the fact, that it's not memory or hardware. It's something in the software. That focuses people faster too.

Like, say it was that ARM ultrabook system. It's trivial then for the 'little guy' doing the application maintenance project out there somewhere, to just buy a new ultrabook system, and prove to himself, that his memory isn't the fault or anything. He's much more likely then to gain enough confidence to send it up the chain.

I find all of this very strange anyhow, because an ultrabook chip, under a license from ARM, doesn't even have to be made on seven nanometer for heaven's sake's. There must be some segment of a team, and some available capacity somewhere in a process generation, that is churning out chipsets or whatever, that it's there and available to push some ultrabook ARM chips down the line too? Anything, that doesn't need to be a Cavium server grade machine either. But yeah, that small caveat of whether it's possible to implement ECC of some kind in the ultrabook product too. I don't know.