"With systemd, setup/startup/stop/teardown responsibilities are concentrated with PID1 and it's helpers.
Before, you'd have the same concept spread into a dozen of different systems, each only doing part of that functionnality."

Which is exactly how it should be.

PID1 only needs a small subset of those capabilities to do its job. And because it is PID1, because everything after has to rely on it, it's essential that it be well behaved and stable. Therefore it is essential that it have only the required set of capabilities and absolutely nothing else should be added or linked to it.

Other things can and should be done by other systems, not concatenated together and poured into PID1 where an error can bring the house down.

To add to the Galaxy Zoo suggestion:
Have a look at this book: "Statistics, data mining and machine learning in astronomy" http://www.britastro.org/journ...
I have my BS in Physics, but I write software. I think it would be pretty hard at 10 hours a week to pick up the math of most of the advanced topics - even mechanics. But, learning statistics and data mining and having public access to data like the Sloan Survey would put you in the position to make real discoveries as an amateur. And, a modicum of competence in statistics and data mining may give you some good options for paying gigs.

"That's because you're a curmudgeon. :)"

Is that the new word for 'adult?'

"My data are important to me. I shouldn't need to buy a server to prevent my data from being corrupted."

But you do nonetheless. My current machine was bought for one reason - price - and lacks it. When I've built my own systems in the past I have always used it. Scoping out parts to build a new one, I see the price of sane memory has only gotten further out of line than I remember. :(

This is one aspect of a market where the buyer does not understand the product well enough to make intelligent choices. If computer buyers understood the technology, at least 70% of them would insist on ECC, and as a result economy of scale would have eliminated the price premium long ago. Instead, manufacturers continue to skimp a few pennies on the RAM by default, creating an economy of scale advantage in the other direction, which only reënforces the bad allocation and ensures it continues.

Instead of ECC memory they should call it 'sanity-checking memory.' Maybe then people would understand what it is enough to realize they want it. But since no one in particular stands to make a windfall by doing it, no one promotes it.

Yep. Still sufficient reason to reduce the amount of plastic that gets into the oceans, but unfortunately, it seems really hard to get people to take any positive environmental steps unless you exaggerate it into ugly, apocalyptic terms. And even then, for every person you convince by it, there will be one who heard that it was exaggerated and concluded that therefore nothing needs to be done at all.

Very often being 'professional' means being willing to do shoddy work in order to conform with a business plan and go home every day at 5 with a paycheck. The amateurs are the ones that sometimes love their craft and insist on doing it right.

I am saying that its design sacrifices robustness in favor of performance and features at every turn. It might be more crashy, but the bigger problem is it ensures you have no usable logs when it crashes. And it doesnt have to be a crash for it to be troublesome, for a single example in the quest for shorter boot times it starts services without making sure that dependencies are actually working - that normally wont cause the entire system to crash but so what?

Still not what I want on my system. I dont really care how long it takes to boot, I just want to make sure that when it's finished it's really finished. Systemd in so many ways copies windows concepts instead - like how they make it supposedly boot faster - by rushing along to draw a GUI before things are actually ready to use.

Not saying systemd is as bad as windows - and the massive improvements in boot speed are not all illusory! but they do come at the cost of reliability and correctness, and that's simply not a good tradeoff for people using the OS in a traditional manner.

"You don't have to. If you really want your old way then just have journald pass everything along to syslog and it's back to normal."

Unfortunately that's not quite true. You *can* configure systemd to spit out text logs as well as the binaries but that is a delayed process, so in the one case where you MOST want text logs (where a crash has occured with the file open) it's absolutely worthless.

I didn't follow the speculation, but perhaps you'd know: did they realize that splitting was an option? Did Boeing and SpaceX each get half a loaf, or did NASA somehow manage to "grow the pie"?

If so, where will they dig up additional billions in funding? If not, will either SpaceX or Boeing be able to accomplish a large fraction of the work for a fraction of the funding they'd hoped to get?

I'm ecstatic to see them say "Why not both?", since if the government is going to be spending tax dollars, I'd rather see it go to a good scientific cause than... well, to a lot of other things that the government is prone to spending money on. But It's a fair bit of money, even in government terms, and I hope it's being spent wisely rather than having a Solomonic decision that gives us two halves of a baby.

I think there is a major difference between having a big possibly over-complicated application program in userspace, and putting something like that in a critical spot in the system itself.

If your application program has a flaw, it's probably not a huge deal. Maybe it crashes occasionally. You save often, you have autosave, it's not a big deal.

But a system component that can crash the system, render it unbootable, hand control to a hostile third party, etc - it's much more important in that case to keep things clean and proper to keep the machine itself stable.

Part of the disconnect between the Sysd cabal and the traditionalists here is about what we mean by the machine. We are often running linux on bare metal as our workstation. From what I have been told, they typically run it in virtual machines on server farms instead, and use Apple workstations. So from their point of view, it is just another application, and it shouldnt be a big deal to restart it occasionally - especially after they put so much work into improving boot times. But from our point of view, we dont care much about fast boot times, we want a stable system that doesnt need to be rebooted all the time.

It's not.

What, you are surprised? You expected accuracy in a slashdot blurb?

Well, actually it's the Supreme Judicial Court, or SJC.

The way this headline is written, it's as if they'd written "Armstrong jumps to moon", and neglected to mention in TFS that he was jumping from the ladder of the lander to the surface. TFS says "managed to 3D print, and assemble an entire automobile", and that's misleading to the point of lying.

It's a cool, impressive, incremental achievement, but they haven't landed on the moon here. And tech reporting, and tech in general, would be better served by accurate reporting of it.

Yep. MOOCs don't serve the important part of the teacher's job. Teaching is best as a dialogue. A videotaped lecture is little different from a book, in that the information is fixed; worse, unlike a book, you don't even get to read at your own pace. It's not without value, since some things adapt well to that and different modes work for different people, but it's still missing the two-way communication that a real teacher provides.

People have pushed MOOCs largely for the learn-a-bunch-of-facts classes, such as science and tech. Technique is also a "fact"; it's stuff that can easily be tested and graded. The things that are missing are the parts that make us consider a student well-rounded: history, literature, sociology, art. These sound trivial to nerds but they're about innovation and communication. They, too, have to be practiced, and it's not something that can be memorized. Even the STEMmest jobs are ultimately about people: seeing what people want, finding ways to tell them your ideas, building up a story together. And that's something that a real teacher can help with, and a videotaped teacher can't. (Nor can a videotaped teacher answer questions or ascertain just why a student isn't "getting it". Even a "great teacher" is little more than an actor when on video.)

Teaching is too often undervalued as if they were just handed a book. It's a skill of its own. We STEM nerds often undervalue that skill because it's not easily graded on a multiple-choice test.

TFS is written primarily in the present tense, which is kind of odd seeing as Lem has been dead for nearly a decade. We are already living in Lem's future, and the future for Lem himself is pretty much a steady-state.