Re: It's almost as if....
"Isn't the point of technology these days to move everything into the cloud (or at the very least, centralize it)? Git sort of threw that philosophy away."
No, let's put this headstanding nonsense back on its feet.
The closer something is to your processor, the faster you can work on it. L1 cache speed is insane. That 8GB of main memory is pretty quick but the round trip will hurt. The hard disk (assuming it's a disk) is the slowest thing *in your machine*, which is why spilling into swap space is a total deal-breaker, makes a program take years instead of minutes so you go buy more RAM.
But the slowest, most unpredictable thing above all-- that's the network.
So you have caching. Local caching on your hard disk of stuff off the network, local caching in main memory of stuff off the hard disk, local caching in the CPU of the stuff in memory. Caches are smaller and faster and more expensive as you get closer to the cores, where all the work has to happen. We get by with using what we absolutely need for the thing or the few things that we're working on *right now*.
Those nasty details you mention? A git repo is a cache, and it makes everything faster. Except a cache miss, which is always expensive, and in this case is also known as "the conditions immediately preceding 'git clone'".
Apparently they made it so your own unpredictable (by the machine) behaviour is what decides the parts of the cache that burn time getting populated. It makes sense. But you pay a different price, since you didn't sync everything at once then you can't get additional unexpectedly needful files unless you're still on the right network, and of course opening such a file still takes a tad bit longer than if it was on disk.
And the Godforsaken cloud... it's what, a big imaginary hard drive on the other side of the network? But it's even slower, and its reliability is every bit as imagined. The people running that way didn't utterly "throw away the philosophy" of getting more data closer to your CPU, because all these silly apps still have to live in *your* RAM and they still have to phone home, and they hopefully know all about syncing and caching and minimum necessary traffic and whatnot.
To answer your question, no. The point of technology these days is NOT to move everything into the cloud or centralize it. The point of technology these days seems to be "make everything operate as much as possible simultaneously, in parallel non-interdependent tasks", and that means decentralizing a great deal.
P.S. clouds are made of dihydrogen monoxide, so how about that?
Biting the hand that feeds IT
