If you could pick any EOL operating system to revive, what would it be?

[TerraHertz]

I've never seen a "corrupt" disk, nor have I ever had any disk problem that could be fixed just because of NTFS. Maybe I'm just lucky, but considering how many disks I've had just keel over and die, I doubt that's the case.

From a personal experience too long to relate, I came to the conclusion that Microsoft's NTFS driver code contains something buried in it that knows how to brick hard disks by issuing a bogus HDD flash firmware update command. And that it's part of some kind of 'detect non-kosher Windows installs' system, with the drive-bricking being one of the punishments it's able to dish out.

Here's some interesting history of the development of NT/2000, by David N. Cutler, NT's chief architect.
http://windowsitpro.com/windows-client/windows-nt-and-vms-rest-story
Windows NT and VMS: The Rest of the Story - Nov 30, 1998.

Something I'll add to that, which I can't recall where I read: another reason NT worked so well, was that when Cutler was negotiating with MS/Gates to come onboard, he demanded a guarantee that MS/Gates would leave him the f*ck alone to get it done right, as opposed to forcing him to include all sorts of stupid Bill Gates bullshit in the OS.
No BOGU(*) for Cutler.


* - Bend Over and Grease Up. (Said when Gates was known to be approaching.) A term that was popular among MS employees for some time, until Gates declared its use would result in instant dismissal.

[AG6QR]

For me, it would be CDC NOS/BE for a CYBER 170 series mainframe, simply because that's where I got my start in programming.  That OS is well beyond EOL, much harder to resurrect than any version of Windows.  The OS is fairly tightly coupled to the machine hardware architecture, and the architecture is quite bizarre if you're used to modern byte-addressable computers with 8-bit characters.  It has a 60 bit word length, 18 bit address space, 6 bit characters (10 characters per word).  Notice that 6 bits is not enough to hold all the upper and lower case letters, digits, and control codes, so the characters are upper case only, with a facility for encoding lower case using 12 bits per character.  The machined was a Cray-designed supercomputer of its day, made to run FORTRAN programs fast.

The filesystem was also a bit odd.  It had no subdirectories.  There were "local files" and "permanent files".  Local files would disappear when you logged off, but programs interacted only with local files.  You had to use the "SAVE" command to copy a local file to a permanent file before you logged off, if you wanted your work to persist.  And you had to use the "GET" command to copy the permanent file(s) back to your local space the next time you logged on.

That's just scratching the surface of the system's idiosyncrasies.  I have a different perspective on computing because I started on that, and moved through IBM mainframe VM/CMS and TSO (which also have some idiosyncrasies of their own), and DEC VMS, before moving to Unix, and finally DOS, Windows, and Mac OS X.  Modern computers and operating systems have largely converged so that they have a tree structured directory, ASCII (and/or Unicode) character set, 32 or 64 bit address space, 8-bit bytes, a hardware stack that makes recursion easy, etc.  My kids will probably never know a computer without these features, and may have a hard time imagining how else things could be done.  But there were once other ways of doing things...

[Whales]

For those thinking about BeOS: have a look at haiku OS

[Cerebus]

... 6 bit characters (10 characters per word).  Notice that 6 bits is not enough to hold all the upper and lower case letters, digits, and control codes, so the characters are upper case only, with a facility for encoding ...

ICL mainframes had a very similar 6 bit character arrangement. Right pain in the bum it was. Worse still, most line printers cost more if you wanted them to handle the full character set so the vast majority were UPPER CASE ONLY WHICH MEANT READING ANY PRINTOUT GOT A BIT TIRING. ESPECIALLY AS YOU COULDN'T SPOT THE START OR END OF SENTENCES.

The filesystem was also a bit odd.  It had no subdirectories.  There were "local files" and "permanent files".  Local files would disappear when you logged off, but programs interacted only with local files.  You had to use the "SAVE" command to copy a local file to a permanent file before you logged off, if you wanted your work to persist.  And you had to use the "GET" command to copy the permanent file(s) back to your local space the next time you logged on.

As best I can tell it was my personal favourite, Multics, that introduced a filesystem that looks like what you'd see nowadays. Prior to that filesystem structures were not merely different between OSes, they were wildly different. And moving files between disparate operating systems usually required the writing of specific conversion programs for both ends not "chuck it on a flash drive".

That's just scratching the surface of the system's idiosyncrasies. 
...
 But there were once other ways of doing things...

What people nowadays may not realize was that at one time it was quite normal for systems to be so wildly incompatible with one another that it was quite normal to rewrite your application code if you merely moved up within a manufacturer's range of machines let alone when moving between machines from different manufacturers.

[StuUK]

ICL mainframes had a very similar 6 bit character arrangement. Right pain in the bum it was. Worse still, most line printers cost more if you wanted them to handle the full character set so the vast majority were UPPER CASE ONLY WHICH MEANT READING ANY PRINTOUT GOT A BIT TIRING. ESPECIALLY AS YOU COULDN'T SPOT THE START OR END OF SENTENCES.

Remember it well, my first job was Cobol on a ICL1900 on coding sheets, punched cards for JCL and the odd fire or two when the friction made the card reader too hot...

[Karel]

If you could pick any EOL operating system to revive, what would it be?

Nothing, I'm quiet happy with what's available today, much better than eol stuff.

[Delta]

MS-DOS 6.2

It used to chug away on the 386 boxes this poor teenager bodged together.  DBLSPACE to compress the data on my tiny HDD, COMPLNK (or something) to transfer files over a parralel cable to another, Telix for BBSing, FastTracker for music making, and I could write a CONFIG.SYS with my eyes closed!

[BravoV]

MS-DOS 6.2

It used to chug away on the 386 boxes this poor teenager bodged together.  DBLSPACE to compress the data on my tiny HDD, COMPLNK (or something) to transfer files over a parralel cable to another, Telix for BBSing, FastTracker for music making, and I could write a CONFIG.SYS with my eyes closed!

Meh .. you missed the upmost important one, squeezing out the last free bytes tweaking out Quarterdeck QEMM386, and CONFIG.SYS alone aint enough, you forgot the AUTOEXEC.BAT. 

[Delta]

AUTOEXEC.BAT goes without saying!

I also had a programme called 2MEG that could squeeze a full 2MB onto a standard 1.44MB 3.5" floppy!  Not tried to read any of those disks for a while though...

[T3sl4co1l]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

[mrpackethead]

windows 10

[Cerebus]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

It was 1.44Mb.  I threw out a huge stack of Windows, Netware and other OS floppies during a clearout a couple of years back.

[rdl]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

I remember I would install DOS first, then use it to copy all the floppy disks to a folder on the hard drive. By doing that you could install Windows from the HDD and if you kept that folder it would look there whenever it needed something instead of telling you to insert disk number whatever.

[Fortran]

It was 1.44Mb.  I threw out a huge stack of Windows, Netware and other OS floppies during a clearout a couple of years back.

Ooh! NetWare was my favourite server around 1990-2000.
It needed a 100MB DOS partition to boot what was basically Linux, I think. Never understood why.
Maybe it wasn't Linux based back then?

[vis1-0n]

I've never seen a "corrupt" disk, nor have I ever had any disk problem that could be fixed just because of NTFS. Maybe I'm just lucky, but considering how many disks I've had just keel over and die, I doubt that's the case.

Have a production bureau producing 200K new files a week. I have had folders disappear, but put it down to user error. It clicked 2 weeks ago when I was deleting the older folders when active files disappeared of the filesystem in Server 2008. It was freaking NTFS itself. Folders went missing in alphabetic order, the more I deleted from the specific folders more folders went missing in its root folder and in alphabetic order.

Of course with my daily backups I ended losing 40% of the current production, undelete utilities helped a bit. NTFS aint bulletproof. Unless it was windows explorer at fault.

[SeanB]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

I remember I would install DOS first, then use it to copy all the floppy disks to a folder on the hard drive. By doing that you could install Windows from the HDD and if you kept that folder it would look there whenever it needed something instead of telling you to insert disk number whatever.

Also worked for the CD version, which was a lovely holgrammed one as well.  Boot from CD, format c:/s and then copy the cabinet files across with the setup.exe and then reboot, and cd to the directory and run setup, answer the 5 questions and let it go for it. Also later slipstreamed a bootable cd with the files and the required inf and files for the boards I mostly used, the video cards and the network drivers, so that it was even faster, and no need to actually install the FDD in the case, saving a few bucks on the build.

Then when I had to build a batch, just clone the drive after doing this, or just run setup till the first reboot and clone it then, saved around an hour per unit, as I could have one built for the second, and then simply clone the drives while doing the hardware, then simply install drive, power on ( and the quick burn in that setup gave by accident) and enter key.  Some were even quicker, used regedit to change the key on the cloned whole drive to match the key stuck on the case.

Found a few of those disks recently clearing up, plus a never used copy of WinME, still in the wrapping.

[eugenenine]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

It was 1.44Mb.  I threw out a huge stack of Windows, Netware and other OS floppies during a clearout a couple of years back.

Same here, tossed all mine floppies when I moved a few years ago as well as old PC's and parts.  My original IBM XT with 10M HDD and working network card, my pair of NT4 domain controllers.

Now if I ever get time to get my Amiga working again I'll have to buy floppies (again).

[C]

WOW, 6 bit characters what a luxury some of you had,
model 28 teletype with five-bit Baudot.
mini computers used as embedded processor, no boot roms, hot patching a running computer.
Then there was the two computer as one thing. Was easer to use a second computer as expanded memory then to create expanded memory. What was left of second cpu became a little smarter io.
Then some just were tubes or transistors.
But think of what happens when you have a RSX11M with users doing some embedded work linked to an embedded PDP-11 & embedded z80

Big cause of "corrupt" disk is a power failure.
NTFS with it's transcription log can handle this better, With FAT your sol if disk was being written to when power fail happened.
Bad programming can kill both.

Most here are thinking of a OS for one user or an OS for many users on one CPU.
I think the real future is an updated TurboDos where each user has a CPU linked to more CPUs in box with linked boxes.
Today some have two or more screens to one PC, what if you had two or more screens to two or more PCs that acted as if it was one PC to user.
An updated TurboDos+RSX11M clone.

[Red Squirrel]

Windows 2000 hands down.

Enhance it under the hood to handle 64-bit, more ram, etc... to be up to today's standards, but keep the user experience exactly the same.

[dannyf]

I was waiting for someone to bring up beos. Had it prevailed over next, the world of computing would have been different today.

[Cerebus]

I was waiting for someone to bring up beos. Had it prevailed over next, the world of computing would have been different today.

BeOS was mentioned in #85. As to the latter, I don't know but I suspect not. BeOS and NeXT innovated in different ways but neither brought anything truly new or remarkable to the mix, and both suffered from a lack of core simplicity and elegance that is, to my mind at least, the mark of a good operating system design.

[TerraHertz]

WOW, 6 bit characters what a luxury some of you had,
model 28 teletype with five-bit Baudot.

This is partly why there's no OS, past or present, that I'd like to see revived. Because *everything* is ASCII-based, even if kludge-extended with Unicode and UTF-8. And it's inadequate.

One of my hobbies is sort-of information technology archeology, with emphasis on identifying points at which major conceptual mistakes were made. Especially when they became permanently bricked in, so we still suffer from them now.
There are a LOT of mistakes and omissions in ASCII, and boy do we suffer. Just most people don't realize how badly, because there are horrible work-arounds for the worst of them, and we're used to these ways of doing things.

So, I verrrry-slowly work on a replacement for ASCII. And some other OS-related stuff based on that rejig.
Knowing what would be possible with some basic fixes to 'accepted crap', makes me very negative about existing OSs.

Big cause of "corrupt" disk is a power failure.
NTFS with it's transcription log can handle this better, With FAT your sol if disk was being written to when power fail happened.

Sole cause of any kind of process-state/storage failure with present computing systems in power-fail situation, is the absence of some REALLY SIMPLE measures at the hardware and OS level. Having designed gambling machine systems that were absolutely immune to any data loss from users playing chopsticks with the mains power switch, I know it can be done at literally no extra cost. It's a design conceptual thing. No it isn't a UPS.
The fact that PCs (and virtually all commercial computers) are not capable of weathering sudden power loss and restarting where they left off, is just pathetic. Unbelievably incompetent design. And it makes them useless for most real world control and state-critical applications.

Most here are thinking of a OS for one user or an OS for many users on one CPU.
I think the real future is an updated TurboDos where each user has a CPU linked to more CPUs in box with linked boxes.
Today some have two or more screens to one PC, what if you had two or more screens to two or more PCs that acted as if it was one PC to user.
An updated TurboDos+RSX11M clone.

Putting it another way, when early computers were expensive it was normal to have many users on one machine. Even with PCs/Windows this was assumed to be a common use case, hence the stupid multi-user pile-of-crap that is one of the reasons all MS Windows versions suck dogs balls. (Combinatorial complexity explosion, and endless security holes.)

You're right that the future is opposite to this. It will be networks of computers owned by individuals. No 'multi-user' tangles. More like 'personal clouds', with extra hardware modules of all kinds hot-plugged to a live networked system as required. The present corporate-owned 'cloud' ideas are kind of resistance to the inevitable, and corporations trying own everything, as usual. They are going to fail.

And there's no OS, past or present, that comes anywhere close to what's needed. Because ASCII is missing some critical concepts.

By the way, for those interested in OS development: http://wiki.osdev.org

[Ampera]

I've never seen a "corrupt" disk, nor have I ever had any disk problem that could be fixed just because of NTFS. Maybe I'm just lucky, but considering how many disks I've had just keel over and die, I doubt that's the case.

Have a production bureau producing 200K new files a week. I have had folders disappear, but put it down to user error. It clicked 2 weeks ago when I was deleting the older folders when active files disappeared of the filesystem in Server 2008. It was freaking NTFS itself. Folders went missing in alphabetic order, the more I deleted from the specific folders more folders went missing in its root folder and in alphabetic order.

Of course with my daily backups I ended losing 40% of the current production, undelete utilities helped a bit. NTFS aint bulletproof. Unless it was windows explorer at fault.

I experienced the same when deleting files from a NAS. Therefore I would blame explorer.exe, instead of ntfs.
FYI, the desktop system is Win7 64, the nas is Ubuntu 15.04.

Look into Ext2FSD. It's a file system translator for windows, you can use the EXT4 file system with writing and reading. It's a tad experimental, but I have never had an issue unless you intend to run windows programs from it.

[Don Hills]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

It was 1.44Mb ...

Some versions of Windows (and OS/2 ad PC-DOS) did use "1.44MB" diskettes formatted to hold 1.68MB and 1.86MB respectively. The Windows version was called DMF and the IBM version XDF. There are Wikipedia pages for both formats.

[Halcyon]

Remember how Win95 was distributed on those 1.76MB (or whatever it was) floppies?

Tim

It was 1.44Mb ...

Some versions of Windows (and OS/2 ad PC-DOS) did use "1.44MB" diskettes formatted to hold 1.68MB and 1.86MB respectively. The Windows version was called DMF and the IBM version XDF. There are Wikipedia pages for both formats.

Yep, DMF used 21 sectors per track. It made copying those disks a little more difficult because the data couldn't fit on a regular 1.44MB formatted floppy disk.