Breaking the nerd internet: Three overlapping generations of tech history – in one selfie Sysinternals founder Mark Russinovich's after-dinner photo just flipped the nerd world into Kardashian-like levels of internet meltdown. Russinovich posted a selfie on his Linkedin page that shows two brilliant OS programmers whose work has shaped the modern world of computing… oh, and next to them, Bill Gates. The star OS …
A while ago I did post a suggestion here that maybe there should be some revised badge rules for these blatant shitposters. No idea if anyone even sees the old discussion forums any more, even though they still work... I can't see any way to navigate to them from the main site.
This post has been deleted by its author
Yeah that's alll well and good, and we are reporting them, but how many strikes is enough? When was the last time he said anything that wasn't just a single line of abuse? He's only here to shit on people, just ban him and make this corner of the internet a slightly better place.
I don't think banning is a reliable solution, given it's easy to just sign up again with a new throwaway email address, and perhaps Tor if the Reg blocks IPs. Better I think to let them keep their identity, but don't give them the veneer of respectability that comes from the coveted silver badge, when almost everything they post gets 10x as many downvotes as upvotes.
It would be some improvement to change the rule for the silver badge to require a net 2000 upvotes, with each downvote subtracting 1 from the total.
Possibly a way around this might be to require serial downvoters* to properly explain WHY they downvote someone else's comment. Their entered text would need to be analysed by something (could this be a good use for AI?) and rejected if it didn't make sense - i.e. to prevent someone mashing the keyboard randomly as a reply. Their reply, if accepted, would also be added to the conversation for all to see (but non-serial downvoters would be allowed to downvote the serial downvoter's reply without needing to explain why). If their reply is rejected, their downvote isn't applied. Too many rejections = suspension from commenting for a period.
* E.g. anyone who has downvoted others more than they have upvoted others in the past, say, three months.
(Hides while waiting for the inevitable downvotes!)
> I still have 2/3rd the hair I had at 15.
Yeah, sure, me too.
The problem is not solely about the amount of hair. It's also about where it grows.
As if under the influence of gravity, a lot of it now seems to grow out of my shoulders, sliding earthwards at about 1cm per year.
Somehow I don't think it's going to be the Finn who take a tumble at the Reichenbach Falls this time. :)
I imagine the meeting might have been rather entertaining as I doubt anything to their conversation touched on their particular professional interests. Their life experience between them would have provided many other interesting topics, I imagine.
.... but he could design an OS from scratch, he didn't need to ape an existing one. And a really bad one - just it was used in most universities and students belived it should have been the best because it was used at universities... so more than 50 years later we are still stuck to an OS designed for teletypes and punched cards.
Exaclty what Torvalds and Stallman did. Those who don't understand the several design flaws of Unix, are condemned to reinvent, it, and that will lead to another poor design, like Linux did. Those who really undestand Unix, stay away from it, and design something different and better. Again, Cutler was right. Unix should be left to the history of OSes, and newer, better OS need to be designed for the future.
Poor design works. Poor design is out there getting its hands dirty and doing stuff. The "better" OSs are perpetually in the future because they can't leave their pristine lab and enter the real world where life is messy.
How does your better OS deal with the PC losering problem?
Go in then do it .... create a "modern" OS, how many have you tried and on how many platforms ?
Now try really try to fly your new shiny with any half decent backward compatibilty ...
Now try that on what i presume you think is a modern OS m$.... hell just try get it to do a half decent update.
How you "fit in" in this industry with a gobshite comment like that baffles me ....
What about yours?
You seem to be implying to have an opinion on OS design you have to have designed and written one?
Maybe it would be better to explain what aspects of Unix you like and consider good design descensions? (Personally I think there's merit in the 'one small tool to do a job well' approach for example).
However, Cutler is no slouch, he's still working today in his 80s on the XBox hypervisor, his influence in OS development isn't zero, which does give his opinions some weight.
There's a long but interesting interview with him on Dave Plumbers youtube channel.
The bit where he goes on about Vista is quite interesting. He didn't approve of the build it and maybe test later approach to Vista, which is why they forked the Windows Server codebase at that point, his attitude was 'You're not bringing that over to server'... it's also why they had a codebase they could re-build upon when they painfully learnt the hard way he was right.
Is it so great that the world runs on 30years of patches onto a student project to re-implement a hack to do text processing on a system that was small enough to run on a discarded bit of kit that was too under powered to run the "proper OS" that was being designed for it ?
It at least shows that fate has a sense of humour
(Proud user of Linux since version 0.99 and SLS on floppies, and owner of a signed copy of The_UNIX-HATERS_Handbook )
Your comment has actually made me thing about potential parallels between software and evolution.
A lot of the time, evolution gets painted as some kind of natural 'aim' for perfection. Whereas it's very much a process 'that'll do' and leaves a lot of inelegant 'designs' in nature (very much putting inverted commas around that word!). If something in a creatures biology doesn't present any massive disadvantage and get it killed more often, it'll probably hang around.
Aren't we just a bunch of patches on top of code that ran something that floated about in a muddy puddle? ;)
Add one letter to VMS.... --> WNT
I was a PhD student on a CERN experiment at around that time. We used micorVAX workstations in my experiment.
Windows NT was going tto run on everything - from your deskside workstation right up to big iron.
NT would have been processor independent - same OS on everything. Including the Alpha.
Microsoft killed the Alpha port if I remember correctly, and we see the results now in Windows being not used in top end supercomputers.
At the time physical scientists thought Unix was for longhairs.
Also worth mentioning the Alpha again - Jon Maddog Hall managing to get an Alpha workstaion from his employers Compaq/DEC and sending it to Linus.
This kicking off the port of Linux to another processor.
> NT would have been processor independent - same OS on everything. Including the Alpha.
It was. At peak, x86-32, Alpha, PowerPC, MIPS, and SPARC, I think. 5 platforms concurrently. (All with x86-16 emulation built in.)
> Microsoft killed the Alpha port if I remember correctly, and we see the results now in Windows being not used in top end supercomputers.
Not really, no.
Compaq bought DEC. HP bought Compaq. HP then sucked up to the corporate teat and killed everything that competed with MS or Intel. Result, Intel now owns Alpha. And killed it.
MS kept WinNT on Alpha alive for longer than any other port:
https://www.theregister.com/2023/05/19/first_64bit_windows/
But it is also important to remember that all released versions of NT on Alpha were 32-bit only.
MS kept Alpha alive until the port on IA64 was viable, and then it kept IA64 alive until the ports on x86-64 and Arm were viable.
I do not know MS' internal definition of "viable" but I suspect "self-hosting" is key.
Now it is x86-64 and Arm64... and I wouldn't be at all surprised if there were internal builds on RISC-V or something.
One of the reasons Linux is used on supercomputers because the kernel devs put in the work required to make it scale to thousands of CPUs, filesystems that scale to huge sizes (not just that you can CREATE a filesystem of huge size but that it works efficiently whether filled with lots of small files or a few really big files) plus all the tools for linking computers together into big computing clusters were open source and developed if not on Linux originally certainly on Unix.
Microsoft didn't make that effort because there's no ROI in making changes to support the tiny number of supercomputers sold/built each year.
Or because it's free, open source and everybody know how to program it ?
I'm not sure the inter-process communication on Unix is all that special, most of the supercomputer tasks used something like MPI - which will run happily on almost anything
Although I'm ancient enough to remember when Crays went Unix and the operators tore their grey beard hair out
IPC is a major weakness of Unix. It requires more use of kernel mode than there should be; it should instead be a basic function of the kernel accessible to userland. But in the 1970s IPC and networking (which is just distributed IPC) weren't so central. Unix was designed for timesharing, to keep processes separate from each other. VMS, at least, had DECnet designed in very nicely, more transparently than Unix/TCP networking ever has been.
This post has been deleted by its author
Unix didn't win against DEC VAX and VMS simply because of Unix's presence in universities. DEC, with PDPs and later VAX running VMS, had a strong presence in university research labs for science and engineering. But through the 1980s DEC fell behind on performance and price, and VMS, while a fine operating system in many respects, lost out to Unix because Unix provided portability across the latest fast hardware from many vendors (most of them no longer around) —as well as a fast cacheing filesystem useful for some data-intensive applications.
Cutler was justifiably dissatisfied with DEC dropping their novel hardware and OS projects; I have no doubt that they could have continued to this day producing desirable hardware and software. But they dropped the ball, and it's understandable if he felt skunnered. They picked-up a little with their Alpha systems, but by then it was too late for them to regain their former glory.
The real story is that DEC squandered $billions on VAX9000 and nearly died.
Ken Olsen always said UNIX was snake oil, backed up by Cutler. They had valid points.
It cancelled the much better PRISM + MICA projects. As a result Cutler & Co naffed off to MS.
DEC then managed to salvage Alpha from the wreckage of PRISM and experienced a brief renaissance.
If DEC axed VAX9000 early, committed to PRISM and MICA, it could have had DEC NT in-house, able to run VMS and POSIX apps side by side on the same 32-bit RISC boxes with its own chips. The entire excursion of the DECstations, DEC MIPS kit and Ultrix never needed to happen at all.
UNIX apps on top of DEC clustering on DEC fast RISC could have been a compelling enough option to save the company.
If the IBM/MS divorce still happened, then doing a deal with IBM adding an OS/2 personality to DEC NT was doable. NT shipped with an OS/2 personality in 1993 as it is.
In 1995 IBM shipped Workplace OS/2 on the same Mach kernel as DEC OSF/1 (later Tru64) and underlying NeXTstep. The point being, the will and the desire were there.
Liam Proven> "The real story is that DEC squandered $billions on VAX9000 and nearly died."
Indeed; arguably it did kill them, slowly. The VAX9000 —the long-awaited vector VAX— had the wrong target market with the wrong implementation at the wrong time. DEC saw its traditional market in science and engineering being eaten by so-called "Unix hot-boxes" —the likes of Convex and Alliant with vector and parallel processing that was comparatively easy to use— while the big-money commercial customers could buy an IBM 3090VF for mainframe capabilities and the option of vector processing. Having abandoned the mainframe market when it dropped the DECsystem-10 (nee PDP-10), DEC no doubt felt they were missing a lucrative trick. The implementation was wrong because they chose ECL-logic when the smart money was on VLSI CMOS. Even when their own people came-up with a single chip that could compete with their 9000, Olsen couldn't believe it was possible. The time was wrong because it took them 4-5 years to deliver something, by which time the science and engineering types had already made the jump from VMS to Unix, and by then also had the additional choice of a growing number of fast, relatively cheap RISC-based systems. At the high end... well no one got fired for buying IBM.
I've heard that DEC ultimately sold a total of about 50 VAX9000s; by the time it launched, Convex and Alliant had sold several hundred machines. Seeing that microprocessors were the way forward, Convex adopted H-P's PA-RISC, and the company was eventually bought by H-P, whereupon the Convex developments became the basis of H-P's lucrative Superdome line. Alliant also switched to a microprocessor, but fatally picked the doomed Intel i860.
Liam proven> "Ken Olsen always said UNIX was snake oil, backed up by Cutler. They had valid points."
To be fair to Olsen, I understand his "snake oil" point was that Unix wasn't a standard that allowed software to run on any Unix machine —unlike VMS where executables could generally be moved from one to another. He was right, but I'm not sure that anyone making the leap thought that anyway. A recompile might be enough for standard Fortran or C, and many of the architectures in the mid-late 1980s were novel enough that some tuning of code was necessary for optimal performance (sometimes as little as changing a few custom compiler directives). System calls, particularly ioctls for things like tape and terminal I/O sometimes had to be tweaked, but overall, porting between different flavours of Unix was easier than porting to or from VMS, if working at a low level.
VMS was also bloody expensive and they played the IBM mainframe game of we want you to upgrade so the maintenance on your existing kit is now 10x
Ironically what finally made us switch was they sold Dec Alphas with NT for half the price of Alphas with VMS. And those could run Linux.
Also being able to have Linux on a laptop, which VMS or Sun couldnt, made a bid difference for traveling scientists.
Funny, when he unveiled Windows NT at Usenix Micro Kernels and Other Architectures Symposium, he said that he didn't know why he was there. WiNT was neither a micro kernel nor a non traditional architecture.
Thus cementing * ['V', 'M', 'S'] + [ 1, 1, 1 ] = [ 'W', 'N', 'T' ].
Quite notable that Cutler's foresight in the design of the PRISM RISC (precursor to ALPHA) to include vector instructions with constant stride or scatter/gather, eventually made its way 20 years later in CISC (eg. Intel AVX), and later in RISC with ARM SVE (eg. A64FX/Fugaku), and now RISC-V (ratified as part of RV23 in 2024).
Truly remarkable and visionary, imho!
This post has been deleted by its author
> Is it only a myth, then, that Bill Gates hand-coded the OS that still runs my TRS-80 Model 100?
Ah, yes, the rebadged Kyocera Kyotronic 85 from Kyoto Ceramics. :-)
https://oldcomputers.net/kc85.html
Specs:
https://www.sinasohn.com/cgi-bin/clascomp/bldhtm.pl?computer=kc85
32kB ROM, base 8kB RAM.
Well, a significant chunk (about 1/4) of the Model 100's is MS BASIC for i8080, and BillG wrote a large chunk of that. So, yes, in part.
The original also had a text editor and a terminal app for use with the dial-up modem. Tandy added a to-do list and an address book.
The names in the ROM are (apparently) Junji Hayashi, Jay Suzuki, and Rick Yamashita. So I guess that they had a hand in it. Suzuki credited BillG, though. If he has one claim to coding chops, it is that apparently he was known for writing exceptionally compact code.
So, yes, fair call. I am not sure how much of an OS the machine had, though.
I own an Amstrad NC100 and a Cambridge Z88, and am very fond of both. Both had a lot more complex firmware and the 40-character wide screen of the KC86/TRS-80 M100 sounded too limiting to me so I never bought one.
The original BASIC that came with the Commodore PET compacted away* a single INC instruction that meant that limited arrays to only 256 items (numbered 0 to 255, of course)! You could allocate above that limit but all that happened was that item 256 was placed in item 0's slot, 257 in 1's slot etc. This was fixed for the "New ROMs" (or BASIC 2.0) - the one where, to Commodore's annoyance when they found out, Microsoft had slipped in an Easter Egg so that the command WAIT6502,x (where x was any non-zero value) would print "MICROSOFT" on the screen x times.
* Probably more like they accidentally forgot to put that INC in the code in the first place!
For anyone interested a pretty good interview with Dave Cutler by Dave Plummer, also an ex-Microsofty, can be found here.
3 hours + but well worth it.
https://www.youtube.com/watch?v=xi1Lq79mLeE
And a similar one with Raymond Chen the oft Reg quoted Windows guru.
https://www.youtube.com/watch?v=-vJQv4rgHYE
Cutler was also responsible for VAXELN, a largely-forgotten realtime operating system for VAX processors written in Pascal (and, initially, supporting only Pascal applications).
It is either suprising, looking back, how many different operating systems were developed, built on very different principles, and how few and disparate are the survivors (z/OS marches on as well as Windows and various Unix variants) or it's surprising how few are being developed now (especially given the huge change in the nature of computing over the years).
It's also worth remembering that many of the arguments that today's Unix enthusiasts use to demonstrate its superiority over its commercial alternatives are very similar to the arguments that yesterday's BSD enthusiasts used against System V. Unix doesn't really have a unitary history and Linux is not Unix, though it can do a very good impression. Given that this group have seen it all before (on multiple occasions) and have a common history of driving forward technology projects that have to meet many disparate requirements, my guess is that they have more in common than might be presupposed.
The story of how BillG defrauded Seattle Computer Company is well known. To those who were there at the time.
The story of how Cutler took a rejected rewrite of VMS and made it the NT kernel is mostly documented in Show Stopper and elsewhere. If you know where to look and heard the backstory from Maynard, MA
And anyone who read the MINIX source code in the late 1980's had absolutely no problem finding their way around the first usable release of Linux in mid 1990's. Everything was so familiar.
Funny how the true originators are rarely remembered. The guys who actually created this stuff. Not the guys who made money from other peoples great ideas.
Well some of us do.
QDOS was written by SCC because DR was having so much problems getting their act together with CP/M-86. The guys in Monterey kept promising to ship something that worked "Real Soon Now" but so much time passed that eventually not just SCC but a bunch of other 8086/88 S-100 board makers roiled their own CP/M for x86. As CP/M was a monitor it was no big deal. Eventually a useable version of CP/M-86 did ship and that's what people used.
The BillG QDOS fraud was the "sub-licensing" clause slipped into the (very long) licensing agreement to license QDOS for the Apple II x86 DOS board MS said they were going to ship. MS had a 8080A/8085 board (cannot remember which) that ran CP/M. MS told SCC it was just until DR shipped something that worked. MS then turned around and sub-licensed QDOS to IBM (with a big markup) after poaching the guy who wrote it from SCC. After agreeing not to.
The final outcome is never mentioned. Years later SCC sued MS for contractual fraud and it was settled out of court with MS paying a very large 8 figure settlement. Settlement buried under NDA's. If I remember correctly InfoWorld had a pretty good article on the lawsuit. Would have been around the mid 1980's.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
