UNIX V4 tape successfully recovered: First ever version of UNIX written in C is running again

Computer History Museum software curator Al Kossow has successfully retrieved the contents of the over-half-a-century old tape found at the University of Utah last month. UNIX V4, the first ever version of the UNIX operating system in which the kernel was written in the then-new C programming language, has been successfully …

[Catch-the-Pigeon]

it's a bit like finding an unfinished record by the beattes which belongs in the hall of fame , doesn't exist at the moment But Denis, Brian, Ken would be first in if it did.

[Sam Shore]

So is it the source code, or the binaries, or both?

[Liam Proven]

> So is it the source code, or the binaries, or both?

Both.

I link to analyses of the code, the comment ratio, etc. in the article -- so that means there is source, right? And I show a screenshot of it running, which means there are working binaries?

[katrinab]

But where the binaries ones that were compiled from the source in the last few weeks / months, or are the the original binaries that were used at the time?

[abend0c4]

I'd say a mixture.

The kernel source is there (some C and some assembler) and the C compiler source and a Snobol III implementation in C. There appears to be a FORTRAN compiler that's binary only. There's assembler source for runoff. There are sources for the various commands in either C or assember: chmod is in C, for example while chown is assembler. There are some binary games.

[Liam Proven]

Thanks for the explanation.

I spent almost the whole day writing the article and trying to fit in background and context, then another hour or more getting it running (on a now-obsolete version of macOS). I did not have time to also dig into the emulated PDP-11, examine the filesystem and get into what's there or not.

[abend0c4]

You did the hard work - I just browsed the tarball!

[Anonymous Coward]

> then another hour or more getting it running

So you did actually got it running yourself?

This is why I like your "old school" reporting. A youngster would have just copypastaed some random tweet saying "it works!"

[steviesteveo]

Hear hear. So much of tech reporting is glossing a supplier's press release. It's so refreshing to see real experiences

[Liam Proven]

> So you did actually got it running yourself?

I did. That window in the 2nd screenshot is running on my own iMac, and it took half a dozen practice runs because the backspace key doesn't work in the emulation, so I started it and ran some commands again and again and again until I could type the commands with the minimum typos.

> This is why I like your "old school" reporting.

Why thank you!

> A youngster would have just copypastaed some random tweet saying "it works!"

Ha. That's what I did on Mastodon. ;-D

https://social.vivaldi.net/deck/@lproven/115757268207746143

I feel the dayjob requires a bit more elbow grease on my part.

[[VtS]Alf]

Backspace

*cough* I am sure you tried ^H (Ctrl-H) for backspace? As it was the terminal backspace.

[Liam Proven]

Re: Backspace

> I am sure you tried ^H (Ctrl-H) for backspace?

I did indeed. And

sync

sync

sync

halt

... To shut it down cleanly.

That didn't work either. ;-)

[Peter Gathercole]

Re: Backspace

That should have been

sync

sync

sync

kill -1 1

This last command sends a hangup signal to init, which drops it back into single user mode, after which you can turn it off.

If it is anything like edition 6, the character to delete the previous character was "#". Remember, it was being used on hard-copy terminals, and they can't actually erase a character on paper. And to kill a whole line was "@".

[Roo]

I really enjoyed firing up V4, and poking about - it's a marvel - and provides some valuable insights into how dennis & ken put it together. I went on to fire up V7 from the same archive - and oh boy things had changed a lot. Thanks for bringing these gems to our attention - much appreciated. :)

[Gene Cash]

The player is not compatible with your web browser

TypeError this.player is null (followed by more that I can't copy/paste)

Nice.

[Liam Proven]

Re: The player is not compatible with your web browser

> The player is not compatible with your web browser

Sorry about that. It's a Peertube video. I wasn't sure if embedding would work -- it does on some OS/browser combinations.

Here's a direct link:

https://exquisite.tube/w/qoHtHzpNXncHwrfqpx31tF

[Anonymous Coward]

Re: The player is not compatible with your web browser

> it does on some OS/browser combinations.

It works here (Fennec, LineageOS)

[Anonymous Coward]

Re: The player is not compatible with your web browser

Just a word of caution...

I recently downloaded the Fenec browser APK

from F-Droid and extracted and decompiled it and found a script in the assets directory that harvested browsing history and cookies and sent it off to an obscure advertising company in Indonesia.

Be careful out there!

[Anonymous Coward]

Re: The player is not compatible with your web browser

Interesting. Can you offer more details? I'll run an APK extraction myself when I get back on the computer

[Anonymous Coward]

Re: The player is not compatible with your web browser

Yes, please do and let me know if you find the script too!

You technically don't need to use a computer to extract and decompile an Android app as F-Droid has an amazing app called APK Editor that lets you decompile and edit any installed app on your device and can also repackage and sign your edited app using the built in signing key or your own Android signing key to install your modified app.

The only issue I found was when trying to decompile the Verizon MDM apps found on my Pixel device would crash the APK Editor which required me to manually extract the app and decompile it with Jadx on my Linux Mint desktop computer.

I'll also download the latest Fennec browser from F-Droid to see if it still contains the script I found and report back.

You can also view any preset low level advanced permissions on any Firefox based browser by exporting the omni.ja file and unzipping it using the Termux android terminal emulator.

BTW, I decompiled the Android TOR browser from the Play Store and everything looked good with the preset advanced settings in the unzipped omni.ja including that TOR disables telemetry and other privacy invasive settings.

You could technically create a user.js or firefox.cfg file based off of TORs preset settings found in the omni.ja to use on your Firefox browser on desktop as well.

Do a DuckDuckGo search for Firefox user.js or firefox.cfg for more information on advanced privacy settings.

Best of luck and happy reverse engineering!

[Martin J Hooper]

So is Linux a direct descendant of this OS?

Cool that the tape has lasted so long without much degradation - Would like to see if there is a article or video on how they reconstructed the missing parts...

[joeldillon]

Define 'direct descendant'. Linux was written from scratch (despite what SCO were trying to say back in the day), it has no UNIX code in it. The BSDs originally did, albeit at long remove from this code, though removing anything AT&T copyrighted was part of the process of creating the ancestor to FreeBSD etc (despite what AT&T were trying to say back in the day).

[IvyKing]

IOW, UNIX is to Linux the way that CP/M is to QDOS/86-DOS/MS-DOS with both Linux and DOS written from scratch to emulate an earlier OS. The earliest fork of UNIX was BSD, which was released with license terms similar to the license that used with the release of SPICE.

[Nerf Herder]

I read somewhere that early DOS developers hired by MS had mostly been Unix developers beforehand, hence why some of the commands and CLI behaviour look familiar.

[Yes Me]

early DOS developers

Shame they got \ and / mixed up.

[Arthur the cat]

Re: early DOS developers

Shame they got \ and / mixed up.

Early DOS didn't have subdirectories and '/' was used to indicate command flags (as in DIR /W), so when subdirectories were introduced they chose '\' for command line stuff as being similar to Unix's path separator. The DOS internal file systems routines would happily accept forward slashes as path separators as well as backslashes.

[Peter Gathercole]

When you say BSD, it depends on what you mean. The early BSD 1 and 2.x tapes were additional tapes that you put onto a Bell Labs. installation. They included some kernel replacement routines, some additions to the kernel that could be compiled in, and a whole bunch of additional commands.

Later BSD tapes were complete OSs, but still based on a Bell Labs/AT&T base derived from Edition 7. It was BSD386 and it's ilk that became supposedly distinct from the AT&T code, but this was subject to dispute and a large court battle that resulted in several BSD variants that really were free of any AT&T code that was not judged freely available.

[Liam Proven]

> what SCO were trying to say back in the day

Always remember: that was the SCO Group.

The SCO Group != the Santa Cruz Operation, the original SCO.

[R Soul]

"So is Linux a direct descendant of this OS?"

Nah. It's a cancerous mutant.

Since Linux shares no DNA/source code with Unix, it can't be a direct descendant. IMO Linux also broke the design principles in UNIX a long time ago and that has further distanced it from the One True OS that Ken and Dennis started ~50 years ago.

UNIX started out as a riposte to a bloated, over-ambitious, all-singing, all-dancing OS that was called Multics. [It proved a simple, highly functional, well-designed OS could run on very modest hardware.] Today, the BSDs are a riposte to another bloated, over-ambitious, all-singing, all-dancing OS. How times have changed.

[Yet Another Anonymous coward]

>Since Linux shares no DNA/source code with Unix, it can't be a direct descendant.

That's like saying 60s rock music wasn't a descendant of the blues because James Brown didn't play any notes on a Rolling Stones album

Linux is a Unix, it took all the original design from this Unix. It's diverged a bit now with Systemd etc but you could sit down in front of tihi and type most of your day-day commands

[Steve Davies 3]

SCO, or its descendents would disagree with you on that

they cited errno.h as similarity between Unix and Linux.

The court case went on for years even after SCO went TITSUP their 'executor' kept filing motions. There are whole websites that documented the whole saga.

[Steve Davies 3]

re: James Brown and the Stones

James Brown did play the same notes as used by the Rolling Stones. The difference is the order in which they were played. Watch the Morecombe and Wise Christmas show with Andre Preview... It is a hoot.

There are only a certain number of notes on the Scale. What differs is the order length and combination of them.

[steelpillow]

In his autobiographical account, Linus says that he had wanted to improve Minix - a UNIX variant/copy used for teaching computer science. But he wasn't allowed access to the IP protected source code, so he decided to brew up something like it for himself. He deliberately made his code compatible with the same shell commands, etc. and over time it got more like a full-blown re-engineered UNIX.

In due course the POSIX standard became a thing. Linus refused to comply with it over something or other, because he declared it to be "a broken standard". ISTR that POSIX eventually saw the wisdom of doing that particular thing his way, and adopted it. It is nowadays possible, with some fiddling, to put together a Linux distro that is POSIX compliant and legally entitled to badge itself a true UNIX. This has led certain Vultures to declare that Linux is now UNIX, somewhat missing the point that 99% of Linux builds are anything but.

I also sometimes wonder how much BSD UNIX code has been copy-pasted into Linux at some point or another, although the GPL forbids doing it the other way around.

So it's kinda like one of those incestuously tangled family trees of hereditary aristocrats and their retainers - how direct do you want the inheritance to be?

[Anonymous Coward]

> In his autobiographical account, Linus says

If you were trying to make me feel old, you managed. I recall reading Torvald's post on Usenet on the matter.

[Bluck Mutter]

point of order

"He deliberately made his code compatible with the same shell commands"

Linus created a kernel, he didn't create the userland...that's why Linux is technically called GNU/Linux.

The userland (all the commands that we in Unix land are familiar with like ls, sed, awk, find, cut, grep etc) came from the GNU project (from the GNU/Herd OS)

Without that pre-existing library of Unix like commands all Linus would have had was a kernel.

So what made it all work with Linus PLUS the work Richard Stallman did for GNU.

Thus Linus and Richard are the fathers of what we see in Linux but history has shunted Richard to the side.

Bluck

[retiredFool]

Re: point of order

Well said. Have an upvote. Technology is a group effort with lots of essential pieces. Lately it seems people think one guy is responsible for entire projects/products.

[steelpillow]

Re: point of order

Userland still needs an API (aka system calls) to talk to the kernel. Linus and Stallman both respected the UNIX system calls, allowing them to join forces to create GNU/Linux distros. Never confuse GNU/Linux with the Linux kernel per se.

[Antron Argaiv]

Re: point of order

As Brooks said, "clearly specified interfaces" are key, set them up and then turn the various groups loose to implement their parts.

[TVU]

That is an excellent question to ask and I would place Linux as in independent emulation of Unix as the Linux kernel does not contain any original Unix source code.

I would also say that it is, along with Minix, is a Unix-like system.

[Yet Another Anonymous coward]

>Linux as in independent emulation of Unix

s/emulation/implementation/g

[Liam Proven]

> direct descendant

Direct, no.

Indirect, yes.

AT&T owned UNIX™. It thought it could extract value by tightly controlling it. It was wrong.

A version developed at the University of California at Berkeley, the Berkely System Distribution, grew the OS into something much more capable. AT&T wouldn't let them sell it, though. However bits the university developed on its own were distributed for free. These made their way into many other OSes, including Windows early on (mid-1990s).

In the end the Uni realised it could just remove all the unmodified AT&T parts and release what was left -- 90% of an OS.

Then the community replaced the missing 10% and bingo, a free Unix, but it couldn't be called Unix, so it was called BSD. Note, like all FOSS (although those terms, "Free Software" and "Open Source", hadn't really been invented yet) it could be sold, and it was.

Snag: they focussed on big kit: minicomputers. They ignored those toy x86 PCs.

The GNU project, meanwhile, had written most of an OS except the kernel. It considered using the BSD kernel but foolishly changed its mind and decided to write its own.

Meantime, a Finnish student was frustrated with his Sinclair QL and got a 386 PC. He ran another free Unix-like called Minix. Snag is it was super limited and you could not distribute improved versions due to the licence terms.

So he wrote his own kernel, using GNU tools and other bits.

Result: Linux. It includes lots of bits developed in parallel for other free Unixes and its design is closely modelled on Unix, but it doesn't use any original code.

It's a ground-up rewrite, not directly binary compatible, but close enough.

Much like MS-DOS started out as a ground-up rewrite of the original CP/M for the Intel 8080 chip, but targeting the newer Intel 8086, because the official CP/M-86 was not yet available.

Both weren't binary compatible, both couldn't read or write the same disks, or run the same binaries, but they were close enough, felt familiar, and it was easy to adapt software from the old one to the newer recreation.

Both ended up totally eclipsing the originals and outselling them a million times over.

[Arthur the cat]

The GNU project, meanwhile, had written most of an OS except the kernel. It considered using the BSD kernel but foolishly changed its mind and decided to write its own.

And still are to this day. :-)

[DrXym]

Linux (disambiguated here to mean a common dist with a Linux kernel + user land) is a Unix lookalike. It is not compiled or derived from BSD or System V source code so technically it is not Unix. But it implements the same functions, commands and concepts as Unix so for all intents and purposes it's Unix even though it isn't.

About 20 years ago SCO claimed they owned Unix and threw sueballs at everyone to try and wipe out Linux but failed hard because the implementation was different. They also lost ownership of Unix in the process to Novell - oops.

Unix lives on in various derivatives. The most mainstream one in existence is Mac OS since it is a BSD derivative. Most *nix systems implement POSIX standards so there is a lot of similarity regardless of how they came into being.

[Peter Gathercole]

The SCO Group and UNIX ownership

The SCO Group did not lose the ownership of UNIX, they never had it! Apparently, they were offered the full ownership by Novell for a price, but they never paid the money. They had the document offering them the ownership, which they tried to assert as meaning they had bought it, but when they could not provide evidence that they had paid the money to complete the deal, the court ruled that the ownership remained with Novell.

What The SCO Group (TSG) had paid for was just the lesser amount that gave them the sales and distribution rights for SVR4 and UnixWare, but that did not allow them to assert control over what the source code could be used for. It also did not give them the rights to revoke the perpetual source and development licenses that people like IBM had purchased from AT&T, which is what they tried to do when they accused IBM of contributing licensed UNIX source code to the Linux code base. It turned out that the code they found in Linux (once they actually revealed which code they were disputing, something they were reluctant to do even when ordered by the court) actually came from UNIX Edition 7, which had been put into the public domain as an Ancient Unix by Caldera before even Novell owned it.

One other side effect of TSGs attempt to assert ownership was that it became apparent that they had not passed the proportion of the distribution fees to Novell that the agreement with them said they would, so them challenging ownership cost them a lot of money they didn't have!

One ironic story I heard was that during discovery for the Linux court case, TSG asked the court to order that they be given the AIX source code to examine. IBM obliged by delivering to them an RS/6000 with the CMVC source control system installed, holding the AIX source. When TSG asked what they were supposed to do with this system, IBM offered to sell them training courses and consultancy on AIX and CMVC to allow them to read the code!

[Peter Gathercole]

Re: The SCO Group and UNIX ownership

Sorry, got that wrong. The Ancient Unix license does not put the code in the public domain, it just grants use under a very, very permissive license!

[Peter Gathercole]

MacOS is not actually a genetic UNIX. The kernel is an evolved Mach microkernel called Darwin, and the command set is BSD from after BSD was made AT&T code free.

But, MacOS is one of the derivatives that pass UNIX branding, although only at UNIX 03. The other current branded variant of UNIX is AIX, and in fact AIX 7.2 TL5 and later is the only UNIX V7 branded OS last time I looked.

So yes. MacOS is a UNIX 03 branded version of UNIX, so not the latest specification, nor a genetic UNIX. AIX is a UNIX V7 branded version of UNIX, as well as being the last genetic UNIX standing! Plus, you can still buy new systems that run AIX.

In theory, UnixWare is still able to be purchased from Xinuos. Last year I downloaded the trial version, but as I couldn't see any pricing information on the Xinuos website, I never installed it, as I didn't want to open myself up to unknown future charges!

[Duncan Macdonald]

Wrong disk size

Removable RK05 disk drives had 2.4MB of storage not 1.5MB. (The fixed RK05F had twice the capacity - 4.8MB as it halved the track to track distance. This could be done as a disk cartridge would only be used in one drive so there was no need to allow for drive to drive variations.)

I remember using RK05 drives back in the late 1970s and early 1980s.

[Gary Stewart]

Now, it has grown into a bloated mess millions of times bigger than the OS which inspired it.

Just like Unix from the early days grew to become more useful over time so has the modern descendant called Linux and the more direct descendants BSDs. They are all much more powerful and useful than the originals and support a range of applications as well as the applications themselves and a vast array of hardware that the predecessors could only dream of. Yeah, there is bloat but the extra functionality, the complexity of modern computing and the cost of the equally complex algorithms that allow them to run efficiently (at least as far as the kernel is concerned) should not be underestimated. I will add that the size of a Linux kernel can vary greatly depending on the needs of the application from very small < 1 MiB for embedded systems to very large ~15 MiB for a full featured distribution.

[R Soul]

Re: Now, it has grown into a bloated mess millions of times bigger than the OS which inspired it.

Hmmmm. If someone put a TCP/IP stack and page-based virtual memory into V7 UNIX - BSD4.[234] comes close - it would kick the arse of any of today's monuments to bloatware^W^W^W Oses.

[keithpeter]

Re: Now, it has grown into a bloated mess millions of times bigger than the OS which inspired it.

Sounds like a good holiday project.

Do it and show us.

[John Brown (no body)]

Re: Now, it has grown into a bloated mess millions of times bigger than the OS which inspired it.

Not forgetting to add in the multitude of security mitigations needed in a modern network connected OS :-)

[Neil Barnes]

It's almost as if it were a Christmas miracle!

One of these --->

for Len Shustak, for recovering the data, and one for Liam for using the subjunctive mood; something rarely seen in computing.

[Yet Another Anonymous coward]

Re: It's almost as if it were a Christmas miracle!

>the subjunctive mood

I'm old enough to have their first album

[Liam Proven]

Re: It's almost as if it were a Christmas miracle!

> Len Shustak, for recovering the data

Al Kossow recovered the data, using software Len Shustek wrote.

[Neil Barnes]

Re: It's almost as if it were a Christmas miracle!

Then he deserves a beer, too. I should improve my reading skills, no beer for me.

[Antron Argaiv]

/usr

I have been told...take this for what it's worth...that "usr" stood for "Unix System Resources"

[MyffyW]

Re: /usr

I too was told that. Unfortunately it was about 20 years ago on the RHEL Engineer course, so I'm not absolutely convinced it's canon.

[Malcolm Weir]

Re: /usr

I would suggest that its not canonical at all. If you look at the general naming policies, abbreviated words were common ("bin", "dev", "etc") and acronyms less so (although not unheard of: "dd" is an acronym for "command and copy" -- true, dat)

[Vincent Manis]

Re: /usr

My understanding is that Unix “dd” was named, and given its obscure syntax, as a joke, an allusion to the IBM OS/360 JCL command DD (define dataset), whose famously obscure syntax with arguments such as VOL=AFF=ABC123 and DISP=(,KEEP) has bedevilled generations of MVT, MVS, and z/OS programmers. Accordingly, I doubt that the name “dd” stands for anything.

[Arthur the cat]

Re: /usr

My understanding is that Unix “dd” was named, and given its obscure syntax, as a joke, an allusion to the IBM OS/360 JCL command DD (define dataset)

My understanding is much the same, except that it wasn't a joke, the original author actually liked JCL's DD. (Ack! Spit!) I heard that 45 years ago when I asked the friend introducing me to Unix V6 why dd was such a non-standard command..

[Antron Argaiv]

Re: /usr

My first thought is that "dd" could stand for "disk dump" or "data dump".

But, I'm probably wrong...or as right as anyone else at this point!

[Yet Another Anonymous coward]

Re: /usr

>dd" is an acronym for "command and copy"

IIRC 'dd' was an "affectionate parody" of the long and complicated command syntax of the commands of certain other operating systems.

[Dwarf]

Re: /usr

I always understood dd to mean one of the following, presumably based on how much success the person had with the command line

disk duplicator

disk dump

data destroyer

[Nerf Herder]

Re: /usr

We called it "disk dump" from my vague memory but Wikipedia says it originally meant "Data Definition": https://en.wikipedia.org/wiki/Dd_%28Unix%29

[desht]

Re: /usr

I'd learnt that it was "convert and copy", but since "cc"was already taken...

[Peter Gathercole]

Re: /usr

The dd utility is Device to Device copy. I think it's documented as such on early man pages.

It's actually such an old command it predates the more normal command syntax of other UNIX commands, mainly because it existed before I/O redirection was added to UNIX. It does much, much more than just copying data from different devices, as it can re-block data to different block sizes, extract blocks from the middle of data streams or files, swap byte order for data from different endian systems, and even convert ASCII to EBCDIC (or vice-versa). Even though it's such an old command, I still use it on a regular basis to do all sorts of things.

I even used it recently to overwrite some bad blocks on a device with unreadable blocks (it was data from a SAN storage system which had not been able to reconstruct all the data after too many disks had block failures). Before writing, the device threw errors when trying to read these blocks. After writing zeros, I did not recover the data, but I could read the blocks back without errors.

[Liam Proven]

Re: /usr

As I said: it's a backronym. Invented later to fit the existing term.

[Anonymous Coward]

Re: /usr

I can also confirm /usr simply stands for "user". I swear I'm really not that old.

[Nerf Herder]

Re: /usr

My first exposure to Unix was BSD4.2 running on a PDP11/70. User access was via traditional green monochrome 80x24 terminals, each connected over an RS232 serial interface. One of my uni projects was designing an RS232 interface box (microprocessor-driven) to add "smart" functions to the terminals. Alas, "smart" terminals in the shape of 80286 PCs arrived a couple of years later, quickly followed by client-server computing, so my box prototype never went anywhere and my chance at fame evaporated (well, OK, it never really existed, but I can dream).

I also remember that all our user "home" directories (this is before people qaintly started calling directories "folders") were in /usr.

[Chaotic Mike]

Re: /usr

>> this is before people qaintly started calling directories "folders

That nice Mr. Jobs has a lot to answer to! Randomly, given the impact mobile devices have had on mental health and society (positive AND negative) if he had known what was going to happen, would he still have released the iPhone?

[John Brown (no body)]

Re: /usr

yes

[isdnip]

Other legends of heritage

Stories I heard, which may or may not have truth to them, say that before Unix was in C, it included a language B, which itself was based on BCPL. TENEX (which became TOPS-20) was written in BCPL, which was as pretty nice totally-untyped language that was popular on DEC hardware, at least in 36-bit-land. Does anyone know if that bit about B was at all true or just made up?

[Liam Proven]

Re: Other legends of heritage

Yes, that's true. The lineage went

CPL -> BCPL -> B -> C

UNIX V3 contained an early C compiler. Some sources claim some of the earlier versions may have used B.

[Yet Another Anonymous coward]

Re: Other legends of heritage

So C++ (or Rust, or Go) should properly be called 'P' ?

[Philo T Farnsworth]

Re: Other legends of heritage

No, C++ was the grade the author got.

Yes, old joke, but I"m an old guy.

[Yes Me]

Re: Other legends of heritage

Ritchie wrote a paper "The Development of the C Language" if you want details.

It's in the archive.

[Vincent Manis]

Re: Other legends of heritage

I believe Tenex was written in PDP-10 assembly language, though BCPL was definitely known (I worked at BBN in 1978, and BCPL had been used there for writing applications for several years). B was essentially a BCPL variant slimmed down to fit in tiny memory and filtered through Ken Thompson's syntax preferences, with some influences from Fortran and PL/I. It was used, I believe, to write several early UNIX commands, but, as the implementation was interpretive, and the language was word-oriented (every value occupied exactly one word), B was unsuitable for the byte-oriented PDP-11, and was never used in the kernel. C added variable-sized values, and was implemented as a compiler to native code, so was therefore a very good fit to the PDP-11, and the kernel.

[Anonymous Coward]

> rather than trying to copy the bytes or sectors from media – in other words, the processed digital data – both readtape and Greaseweazle sample and record the raw magnetic flux variations

Thanks for the explanation, Liam. I'd forgot those tools existed.

[Yet Another Anonymous coward]

In astronomy, where you might be using observational data from 1000 BCE, somebody thought of this and made the standard for astronomical image data defines the very layout of data on the tape - just in case any of these new fangled operating systems weren't around in another 3000years.

[Vincent Manis]

Using stock photos is fine, but this one seems singularly inappropriate. it appears to show a set of IBM 729 tape drives and an IBM card reader (2540?). None of these were capable of being attached to the PDP-7 or PDP-11 that dmr and ken used for developing Unix. There is a famous photograph of the two of them at work (I found it at https://www.historyofinformation.com/image.php?id=6430), which would have been much better.

[Philo T Farnsworth]

The card reader/punch is an IBM 1402¹ -- I knew it well, having loaded more than my share of decks into the beast while operating an IBM 1620.² You're right about the tape drives, they are indeed Model 729.³

Unless I miss my guess, the installation is part of an IBM 1401 system,⁴ which, like the 1620, had its quirks. including alphanumeric memory organization based around the venerable 80 column punched card.

The 1402 and its printer cousin, the IBM 1403,⁵ lasted well into the 1980s, long after the 1401 was superseded by the System/360 and System/370 since they were built like tanks and ran forever, pretty much maintenance free (at least in my experience).

Say what you will about IBM's sales practices back in the 1950s and 1960s (and the Justice Department had plenty to say) but their hardware was solid. There was more than a little truth to the old saw that "Nobody ever got fired for buying IBM."

______________________

¹ IBM 1402

² A machine with an architecture so bizarre that it couldn't even add or multiply without loading lookup tables first (earning it the nickname "CADET" -- Can't Add, Doesn't Even Try. See here for Dijkstra's take on the machine.

³ Wikipedia: IBM 729

⁴ Wikipedia: IBM 1401

⁵ Wikipedia: IBM 1403

[Vincent Manis]

You are absolutely correct about the card reader model. I operated an IBM 7044 system for a while, and remember the 1402 well now, as we had one of those, along with the 729s and a 1403 printer. The printer was magical, as its page definition was via a paper tape, and if the tape broke, the next eject operation would spew a boxful of blank paper into the output tray. Before I arrived there, the system was an IBM 1620, which arrived shortly before a scheduled open house. The Computing Centre wanted to show this off, so they wrote some sort of demo program. On the day of the open house, they pressed the Clear Memory button and loaded the program. It didn't work, because they hadn't provided replacement arithmetic tables.

[Philo T Farnsworth]

You didn't need a broken paper tape, just a bunged up FORMAT statement with a '1' in column 1.

Since I haven't touched FORTRAN in several decades, I'm not sure whether they still exist but at least through FORTRAN 9x, output destined for a line printer used special printer control characters, which were not printed, in the first character of each line. A space meant just print the line as usual, a '0' meant double space, and '1' meant eject the current page and print the line on a new page.

I think '+' meant something, too, but memory mercifully fades.

A programming error (or sometimes plain nerdy malice) could empty a box of paper on a 1403 in about ten seconds (and sometimes cause a horrendous paper jam).

Oh, and they were loud. I wouldn't work in a room full of them without ear protection.

On an IBM 3800 printer, their first high speed laser printer (circa 1975), the paper traveled at something like 20 miles per hour (about 32 KPH for everyone else, except those reading in Myanmar), so that kind of error was even more fun.

[J.G.Harston]

As I pointed out in the previous article, it's the first version that consolodates exceptions into the one signal() call, and has the indir() call. So, it's possible that my code that runs happily on v5 will actually run on it, as checking the syscall list there's nothing my code uses in v5 that is missing in v4. :)

That's my Christmas project sorted.

[J.G.Harston]

YAY!!!! Worked first time!

[Anonymous Coward]

Now you've got to find something else to do at Christmas...

[Liam Proven]

> YAY!!!! Worked first time!

Amazing work, JG. My hat's off to you. (Now I make the connection to FB...)

[Anonymous Coward]

"Today, it's part of Unix lore that there's an important functional distinction between the binaries in the root directory and those kept under the /usr tree"

I'll disagree on this. Solaris, the finest SysVR4 there ever was, merged those more than 20 years ago. It's the Linux world that has had issues letting go of traditions it incorporated from an earlier age. Even though as a wholly new implementation, it really never needed them.

[kmorwath]

We already knew...

... that Unix was never designed, just cobbled together somehow, on primitive hardware.

The issue is that lame design still survives. I understand people like zombies, but it's better to kill them.

[zcomputerwiz]

It's a small world

I know someone who worked at AT&T Bell Labs in the 1980s and was involved in developing SCSI controller firmware as well as porting minimal, stripped-down versions of required utilities and producing a shell-script-based floppy disk installer for one of their new Unix systems. They weren’t certain which Bell Labs Unix variant it was when I asked, but they were quite sure it was written in C.

They also shared an anecdote about a call from some thoroughly befuddled Bell Labs programmers working on other parts of the project who were hoping to get help from Dennis Ritchie himself. After they described their hardware, their code, and the strange behavior they were encountering, Dennis reportedly replied that the problem was “intuitively obvious” and promptly ended the call.

Of course, it wasn’t obvious to them at the time, but they eventually realized the issue wasn’t the code at all. It was their overheating hardware, which they were attempting to cool manually by spraying it with phase-change coolant.

[Peter Gathercole]

Re: It's a small world

By the time Bell Labs. UNIX reached Edition 4, it was mostly written in C. Edition 7, which introduced the Portable C Compiler almost managed to eliminate all of the rest of the assembler code, with only a few dozen lines of assembler left in to perform the initial system setup of certain data structures and mode registers before switching to code written in C for the rest of the run of the OS.

[Anonymous Coward]

Toybox

Toybox came in handy recently when trying to run the Android terminal emulator Termux.

Because my device has an older kernel Termux threw an error when trying to "ls" or "stat" but I found a workaround by creating a .bashrc file with: alias la='toybox ls -lA'.

And another alias for 'toybox stat'.

Worked like a charm.

[Anonymous Coward]

Ugh they rewrote in C...what was wrong with assembly? Developers always want to switch to the "new shiny"...what's next? Rust? /s

[John Brown (no body)]

Rust isn't shiny :-)

[Anonymous Coward]

The DEC RK05

held a massive 2.4Mb...

https://en.wikipedia.org/wiki/RK05

I used one on a PDP/11-40 for my degree project in 1974/75. That ran the DEC DOS OS V8. Code was archived onto Paper Tape. I still have the last archive I made before I graduated. All 1246 lines of Fortran. Those were the days. That code now sits on my PDP-11/83. I don't have the hardware that it compiled code for but when I look at it, I see just how bad my coding was in those days.

[Softsuits]

RISC & Apple drove POSIX

Those to cave dwellers scratchings on the wall were so plagiarized! I still in awe of pipes and sed. Sniff. Now I'm getting all emotional.

[samam95124]

Read Lion's book on Unix. It is still in print. The most striking thing about the code is it uses tricks in C that are unbelievable, and would probably be flagged with lots of warnings today.

[MonkeyJuice]

It's the bleeding edge of the 1970s, you put your initials in SIXBIT in the right word of a POPJ instruction like everyone else.

[DartfordMan]

I don't see any mention of that cul-de-sac of Unix ... Xenix. I used it around '83 I think.

[DartfordMan]

Xenix

I don't see any mention of that cul-de-sac of Unix ... Xenix. I used it around '83 I think.

[Liam Proven]

Re: Xenix

> https://www.theregister.com/2025/12/23/unix_v4_tape_successfully_recovered/

You might enjoy the historical context about Xenix that Charlie Stross documented in a comment on my tech blog the other day:

https://liam-on-linux.dreamwidth.org/97149.html?thread=62589#cmt62589

[nxHacker]

If anybody wants to have a play around with it I created a website where you can get a live terminal of it in your browser: https://unixv4.dev

Check it out (and let me know if you have any feedback)

[Liam Proven]

Here's a video interview from local Utah TV. Good to put faces and voices to the names.

https://www.youtube.com/watch?v=m-3RJaKcw_4