Is it a lost cause?

That's probably _why_ they did so well. Management may be able to do the vision thing okay, but things work best when they keep out of the day-to-day.

Presumably, permanent employees.

Sincerely,

Gene Wirchenko

You are painting too rosy picture. Recently I compiled gcc from sources to get cross compiler targeting ARM, more precisely Cortex-M. Using natural arguments to configure I got compiler that worked on simple programs but more complicated ones were crashing. After some debugging I realized that one of compiler support routines were assembled as 32-bit ARM code, but Cortex-M only supports Thumb so processor were crashing on invalid opcode. After some fiddling with configure I found options which resulted in correct Thumb code. You may call this user error (apparently problem is known to gcc developers and they do not think there is anything to fix), but such problems are almost impossible to resolve without debugging at machine code level.

More generally gcc has quite a lot of bugs. It is quite possible that all tricky fragments of your programs are in gcc bug database so gcc developers make sure they compile correctly. But I have seen enough gcc bugs to have somewhat limited trust. If you manage to find new uses not covered by existing bug database (not easy since database is large), then you are likely to hit bugs.

And there is a lot of development effort spent on gcc. Most of my coding is with different (not C) compilers and then hitting bugs is much more likely.

Yes. Despite its kludgieness it did have the advantage of not being a fixed-size stack. The usual method was to do a GETMAIN for the savearea at the start of the called routine - essentially a malloc call. Then it would be freed at routine exit. It _could_ have been allocated in static data for the routine, but this would prevent recursion or reentrancy so I never did it this way.

Thanks, it sounds worth a look.

... but counterbalanced by better phosphor, higher quality CRT and better quality electronics.

thought I got a real bargain - and I would have checked prices before

A quick look shows that 'remanufactured' Wyse 160s with keyboard are

As long as nothing started overwriting the kernel.

The per desktop cost of computing today is as low as you want it to be. A Raspberry Pi has plenty of power for office work and costs less than a decent keyboard and way less than a decent monitor.

The source I quoted above was offering all Wyse terminal models from 30

ROF,L.

You don't have a _clue_ who you're addressing, do you?

Is that the ex-Tarmac 2966 at TMOC?

Is there any particular time that some of them would be there?

I never saw VME/K, but quite liked VME/B once I'd used it long enough to be able to guess the names of uncommonly used commands (hurray for naming consistency!). I also used IBM's OS/400 quite a bit, enough to realise there were interesting parallels between the two. Both had rather horrid semi-hierarchic filing systems, very consistent command names and both had good command search facilities and nice full-screen parameter prompts once you'd found the command you needed. However, where VME/B had nice, long readable command names backed up by an equally consistent abbreviations catalog, OS/400 insisted on 9 character file and command names, abbreviated by necessity and often looking like line noise. For instance, while VME/B called the COBOL compiler COBOL, OS/400 called it CRTCBLPGM (the PL/I compiler was CRTPLIPGM and the RPG one was CRTRPGPGM)

- get the pattern?.

Yes, but Linux seems to have separate limits for the stack and heap as evidenced by ulimit output.

i used these then, lathes with sixfoot swings and mills with beds you could lie down on. But the cleverest thing i thought was the temperature control, all analog.

And those limits are flexible and can be turned off. The stack limit is usually high enough that only poor programs will ever hit it.

It was a hacked up 'mini VME/B' for the machines without the OCP power to run B! Unreliable and poorly supported. The final straw was when the 'next' version was flagged to be a major 're-engineering' that removed management facilities we deemed essential.

We were getting a combined hardware/software MTBF (rolling 13 week average) of just 20 hours. Moving to the home-grown system gave improved facilities, 50% more throughput, and the MTBF rose to about 2000 hours!

I do, although there were some like that within ICL too. VME/B was just big and inefficient, with pretty terrible facilities for our environment. Originally, we had a single OCP 2960, and it could never have run B anyway.

More accurately, then....an alloca call.

It is still the same space. There is a single block of (virtual) memory in which the heap grows up and the stack grows down, but you can define for heap and stack separately how far they allowed to grow.

Typically the stack is limited to 8MB and the heap is unlimited. (there still is a limitation based on memory available in the system and memory used by other processes)

You totally misunderstand me.

In the commercial world, offices running IT kit have an avvpunting rem that tries to calculate the cost of the employee in wages and NI, the cost of the physical space he occupies including things like aircon and a janitor, and the cost of putting a suitable worlkstation on that desktop.

A Raspeberry PI doesn't come under 'suitable workstation;'

At one point, a Windows equipped PC, with software licenses, technical support to fix it every other day, and depreciation was reckoned to cost

There has been some recent surveys showing that the large organisations that HAVE embraced Linux for the desktop has been doing extremely well thereby. Also financially.

Some of the US brokerage houses are showcases.

Yes, a raspberry pi is too lame; except where you just need a plain terminal and forme entry web browser for some corporate stuff.

But look at the higher end SOCs, like the odroid. The xu4 is about par with an intel i5, and it costs $70. Plus a micro-sd card ($40-$120 depending on size and speed), keyboard, mouse and video.

You could get a very workable computer for $250. You can reinstall the OS and apps by swapping the SD card out. Making a master SD card for a corporate system is pretty easy; and enforce source control for all documents. We have the tools, had them for decades. Now is the time to let the general office worker use them.

And it can work on DC from a battery bank easily too, giving you and instant UPS.

I am equipping a NOC room with these, 30 xu4s and the same amount of screens, simple USB switches for 4 keyboards+mice. The screens are 65% of the total cost.

Or you can buy the chromebooks and du a guest OS install. They are around $220 to $450, everything inclusive.

And when it comes to applications needing windows; I have now used Linux/BSD/QNX for more than a decade and have consulted intensively with around 10 different organisations. None have required any windows apps; except on servers accessed by web.

I have even been appinted to make the reference 'word' and 'excel' documents, because the open source apps make them very much more compatible with all windows/macos etc versions.

-- mrr