Does anyone here know much about the Hurd?

I know it got stuck in "which microkernel shall we use" hell for the longest time. They seem to have settled, but it's not clear if the new one is a modern high performance one (under the Mach name), or if they just settled on the older one and suffered a performance hit.

Also, why is a microkernel OS so apparently difficult to construct?

As far as I can see, the basic bits of hurd are all in place: the things that make it an operating system that actually works. But what took it so long? Micro kernel based things sound like they ought to be easier to develop (segfaults instead of a lockup, for instance), but apparently they are not.

Anyone got any experience?

On the other hand, not having a stocked kitchen to some extent makes the place really miserable.

At a minimum standard tea+milk and coffee needs to be provided. Every single co-working place I've worked has provided that and they have zero incentive to make me work long hours.

The best co-working place I had had tea, a coffee machine with an inbuilt grinder (they provided beans too), a biscuit barrel and free printing within reason.

The thing is that providing some free stuff is really worthwhile if that stuff is disproportionally faffy for individuals to provide.

Consoles are great not because they have good specs (they mostly have shitty ones).

The Cell proessor did boast a quite astonishing peak floating point throughput when it debuted. Whether it was better is left as an exercise for the poor sods who had to program the blighter, but in some domains it had much better performance than the top end PCs at the time.

Not sure what you mean about Blu-ray being obsolete

He believes that everyone has a fast internet connection, I suppose. He's forgotten that even people London (e.g. me) have to suffer on a 4mbps DSL connection due to a mix of non upgraded exchanges and very strict planning rules blocking visible RCU boxes.

On a 32 bit machine it is trycky to go over 2GB per process since you suddenly have to become awfully careful about pointer differences.

The Sun Niagara is (or was?) a very differnet beast to the type of SMT on Intel and the Power chips.

AFAICT the Niagara is basically a barrel processor and switches threads once per clock tick. That provides excellent latency hiding, but has nearly a duplicated set of everything for every thread. But they can be smaller, simpler and slower units since they only go active once every N clock cycles.

On the other hand, the SMT of the Intel and Power variety is different. It has one set of stuff and just tries to jam two (or 4) threads worth of instructions into it. I say "just", but of course it's a rather harder problem.

The difference is that the barrel threads go 1/N times as fast but are N times as latency tolerant (giving theoretically a higher throughput). Whereas on intel, the speedup depends on the workload very much more.

The bulldozer is much more like the Intel one than the Sun one in implementation. But the shared chunk is much smaller, and much less sharing occurs between threads.

as able to adapt to the Pb-free initiative without all the drama and years of failed hardware.

Not at all. Mission critical things were excluded, so aerospace, medical and military are still allowed to use lead if they wish.

Non Pb solder is much harder to work with. Try it: it's obvious even using the stuff.

However the problems don't end there. It's done hotter, which leads to more component failure. It also leads to larger thremal stresses due to gerater thermal expansion during soldering. It's harder and less malleable, which means that it cracks more easily. Eutectic solder, and in fact pure lead anneal and indeed creep at room temperature. This means that stress induced damage actually slowly self heals over time quite astonishingly.

Then there's the trouble with tin whiskers.

Pb free solder is a pain in the ass, but not as much as heavy metal poisioning. But make no mistake it is not as easy to work with.

You can't fix sintered metal by annealing it.

Yes, I was garbling slightly. For injection moulds the problem is that they lack sufficient hardness and the drag of the plastic wears out mild steel ones quite fast.

They need to be strong enough to withstand the plastic injection pressure but need to be hard enough to churn out parts without wearing out.

But until someone young with a new paradigm for translating human intentions to machine code comes around, we must stick to all that context free grammar stuff.

So, C++ is the new paradigm? :)

(yes I like C++ and yes I concede it's grammar is not the nicesest out there)

Scheme allows you to overlay pretty much anything onto its syntax.

That, fundementally is the problem with Scheme and Lisp.

Syntax is purely an aid to human understanding. If you see for, do, while, etc you already have an idea of what't going on. In LISP and Scheme, much as in machine code you have a bunch of code (or is it data! no! it's both!) which the computer can churn through. In order to fully understand it, you have to read everything leading up to that point.

OK, that's hyperbole, but I hope you get the point.

The thing is you can overlay any syntax, and people do, because humans love syntax. The trouble is in scheme is that everyone does it differently. It's like C++ classes versus C classes. In C you can make them just fine with nested structs and function pointers. But everyone did it differently and reasing someone elses code is hard. Is this a class? Oh... looks like it, now how do I set it up. This construct function, that macro? In C++ they're all the same and you know what's going on.

My own company announced that they would give each employee $1,500 to kit out their work area however they like. Extra monitor, some nice plants, miniature trebuchet, whatever.

That is what I call appreciation.