But there's still more than two players, which makes it not a duality.

A non-paywalled link (you'll need to scroll down to find the article).

You're welcome!

Yes, it's always better to buy a corpse of a company after the engineers have burned their notes and thrown their desk chairs through the window on their way downtown to mortgage their car and sign up for food stamps, than to buy a company that is not yet a corpse and still has productive and creative people.

Never talk to the police except through your lawyer

...especially when it involves a pressure cooker.

The gold standard, IIRC, was for apps to get launched and be able to do something in 2 seconds or less under Dalvik in common use.

You're waiting between 300 and 600 seconds, one time, and still waiting for apps to launch (which I'm sure is not instantaneous with ART, no matter your hardware).

My daily-driver is a Galaxy S5. It is, by most measures, a rather fast phone in terms of internal storage speed and CPU grunt. I don't notice much (any, really) difference in loading apps between 4.4.4 (Dalvik) and 5.0 (ART) on this device. But I do notice the "Android is upgrading...." process taking much longer with ART than Dalvik when tinkering, as I also did when experimenting with ART on the Bionic before Lollipop made ART a default thing and the Bionic was my daily.

Also, too: You seem to dismiss the notion that Dalvik is not exactly JIT: It is absolutely JIT on first load, and then largely-compiled apps get stored in the (not-accidentally-named) Dalvik cache for quick loading later.

I'm not sure that we, given our combined experiences, can say which one is better. It'd be awesome if I could find a Google study of actual performance instead a Google proclamation of things to come, but they're Google and therefore always in a forward-thinking perpetual beta with no rearview mirror: Whatsa behind me, is not important.

I did know that. I just didn't spell it out. (How do you propose to know what I know, anyway? Preposterous!)

Look, do you want PuTTY binaries directly from the official site, or not? Because that's the problem being discussed.

Or at least it was until you showed up.

You say that, but...

I'm currently toying with ROMs on an old Motorola Bionic, which serves as the central music player for the stereo in my garage.

I've waited minutes for ART to do its thing on a singular package, but I've never waited minutes for Dalvik to launch the same thing thing on a properly-working device using Dalvik.

They are therefore not the same process. You might think that they should be, and I might agree that I think that they should be, but they're simply not -- or at least, the ART compilation process always takes longer than the Dalvik JIT process.

(Also, you sound a lot like the folks responding to those who question memory management on Android: "It's taken care of automatically," they say. "You can't do anything to improve it," they further proclaim. But there's a dozen or so tweakable parameters, and nobody can tell me that they're perfect for my usage out-of-the-box than anyone can tell me that my "swappiness" parameter on a Linux box is sane, out-of-the-box, for what I'm doing with it. It may be a reasonable default, or it may not be. But toeing the line unquestionably on the published Truth Of The Matter, and Accepting that Truth for What Is is a foolhardy thing to be doing. It may also be the fact that this same Bionic had a long string of bad Dalvik VM parameters that lasted for months or years, resulting in general sluggishness and frustration, though the 12.x Lollipop-based line seems better for reasons I haven't yet bothered to investigate.)

The issue is not, AFAICT, that the pictures may have been used for a yearbook: I'll sign up for that being fair use any day of the week. (I mean, sheesh: Where's the fight there, on either side?)

The issue is that the principal threatened a student with a bogus raft full of bullshit (including "reporting to the IRS") over a display of some 4,000 pictures he'd taken of school activities using school equipment, and subsequently displayed on a non-school website.

Similarly, if I create a chalk drawing of a foot or a bicep or a car tire in art class, using school materials and equipment time and instruction, I expect that this drawing will be mine at the end of the day.

If someone wants to use a copy of it for their own work, that's a slightly different matter.... but it will always remain mine unless otherwise and explicitly negotiated and agreed upon.

Where is the line between having a spirited discussion on how an administrator can lack common sense...and providing their contact details (regardless of how public they may be) to a public forum?

I believe such a line is defined in post #49746421.

Either way, it already was public information. There is no such thing as information being more public or less public -- it either is public information, or it is not.

That already-public information might be more-popularized in one instant than it was in the last instant doesn't seem to matter much.

Right. I know this. It's also available from that main site, directly, via anonymous FTP....which, AFAICT, is still something that Windows browsers transparently support just as they always have.

So it's better to compile things millions* of times (once for each low-speed, battery-powered, handheld device), than it is to compile things hundreds** of times on a distributed and efficient server farm that certainly has a few clock cycles to spare.

Am I with you so far? Are you with me?

Good.

Can we compute the increased carbon emissions of this, including wear-and-tear due to increased battery aging and decreased lifespan (which more and more means death to the entire device)? Man-hours wasted staring, waiting for devices to compile their own apps?

Because if we did, we'd end up with a small number for each individual user. But when we multiply out by millions, very small numbers turn into very real numbers consisting of man-days, man-months, and man-years.

For the sake of pulling numbers out of our ass, suppose that there are a million Lollipop users. If each user has a minute of their life wasted waiting for ART to compile stuff (which is a low number), then we're at 1.9 man-years wasted because targeting a hundred platforms is hard, and stuff.

Before long, we'll be at a point where entire human lifetimes have been wasted waiting for ART to make things fast, because -- somehow -- keeping track of architectures and compiler flags is something that Google can't be bothered to do (even though they already know, by way of licensing, all that is needed to know for devices participating in their OEM Play ecosystem).

*: That's obvious; Lollipop/ART, while not hugely common in the installed base, is still quite a hot item on devices that support it/it supports (chicken, meet egg). Millions seems appropriate.
**: A quick Google search tells me that in August of 2014, there were over 18,000 distinct Android device types, but the number of them running Lollipop is at best in the hundreds.

That you're both pedantic and wrong?

You, sir, win the award for Conceded Asshole of the Day.

You also seem to be unaware that floppy disks vary hugely in size. Common sizes on PC hardware varied from 360 KiB up to 1.44MiB. Obviously, one would need three of the former to hold as much as one of the latter. Perhaps you're trying to imply that what I said was incorrect.

Oh, you poor, poor pedant. (3*360KiB)!=1.44MiB.

If you can't get your arithmetic right, how are we to believe anything else you have to say?

That is greenend.org.uk.

Because that's where PuTTY comes from, you dolt.