Thanks for the detailed hints. A few of my own notes follow.

> My first decept comes when you say that you have assigned 16Mb ram to the thing.

In my defense, this *was* a weird V7 hybrid port for modern x86 boxes that were produced decades after V7. Erring on the side of caution instead of fighting obtuse OOM issues seemed reasonable to me.

> Had you run it in simh, you’d get that a pdp-11/45 holds for an amount of around 256Kb max, and a virtual memory space of 64Kb text, and 64Kb data/stack. This probably makes vi completely useless, so you’d finnally end working with ed.

You’ll find that that I moved to V7 on SIMH in the follow-up posts (https://virtuallyfun.com/wordpress/2018/01/17/teaching_an_almost_40-year_old_unix_about_backspace/ and https://virtuallyfun.com/wordpress/2018/01/28/date-and-time-on-unix-v7/), though mainly because V7/x86 had some strange behavior in various places, and I didn’t quite feel like trying to guess if it’s the compiler, the kernel or the application code not living up to the compiler’s standards and assumptions.

> ed has a problem at first.

ed(1) is definitely an experience. I’ve used it in the “Date and Time on UNIX V7” post to write tsdate and its associated man page. You may also enjoy my subsequent attempt at version control with ed(1), diff(1), awk(1), sed(1) and sh(1) (https://github.com/xorhash/fh).

> You say the interrupt key cannot be adjusted. You are right, but a little digging will show you that it is implemented in the kernel, so a short patch to stty(1) command will allow you to solve the problem

I’ve actually taken a way-too-deep-a-dive with backspace on V7 with SIMH, see the first follow-up post linked above. I do believe your patch to be the more sensible solution, however.

> Well, the copy I’m running on simh allows me to write tapes, and I wrote a simple program to convert the tape format of simh (fully documented on the simh documentation, which, by the way, is misteriously written in Microsoft Word format )

I rather dislike reading the SIMH documentation for that reason. Incidentally, I think you wouldn’t have needed to write your own; as reader DOS pointed out elsewhere (https://virtuallyfun.com/wordpress/2018/01/17/teaching_an_almost_40-year_old_unix_about_backspace/), pre-existing tools exist buried in the TUHS archive.

> The version I’m running is a Dennis Ritchie prepared tape, found in minnie.tuhs.org, under the name unix_v7_rl.dsk, I can provide you a full compliant simh configuration file and the disk images I use (already patched if you wish) As today, they are under Caldera License (and not the Open Group) but the Open Group didn’t exist at the date of release of V7 (1978)

With all due respect, you appear to be oversimplifying the legal situation surrounding Ancient UNIX code, possibly also confusing trademark and copyright law.

While, yes, The Open Group did not exist back then (but I had to update the trademark owner for legal reasons; though the “This is wrong.” in the article probably doesn’t properly convey what I meant to say, cf. the conversation above this one), but that’s irrelevant for *copyright*. The Open Group has no copyright on UNIX code, never did.

See also https://virtuallyfun.com/wordpress/2018/11/26/why-bsd-os-is-the-best-candidate-for-being-the-only-tested-legally-open-unix/ about that.

> Another thing you need to patch is the date(1) command.

In my “Date and Time on UNIX V7” post, I actually went out of my way to avoid touching date(1) at all in favor of a completely automated solution involving %UTIME% in SIMH boot.ini.

> Well, this was my two cents contribution. Now you have for a next part blog.

Which, uh, already exists. But thanks for the pointers!

> I have also patched getty.c to allow for /etc/issue showing at your console, and LOGIN and USER environment variables defined in login.c

Those are very sensible changes to make if you want to backport BSD software from around that era, agreed.

> Your TERM variable, although has to wait to be initialized in your .profile file, but trying to fit vi(1) in 64Kb+64Kb is just a pain (i will probably do that someday, but for the moment I prefer to export this to my host and use git to maintain source code versions

[insert rant about how BitKeeper keeping the SCCS weave is the correct thing to do here]

> For now I’m trying to be able to compile a kernel version adapted to use larger disks (rp) than rl disks (max 10Mb) so I can fit all the system plus sources in a single disk and use rl disks for swapping.

Best of luck!

> If you need some support, don’t hesitate to ask me

Much appreciated, thank you. I may get back to you eventually!

Nope, you need a patched stty.c, as the kernel has already implemented support to change it, but at the V7 edition, It had not reached userland still.

V7 sources are owned today by Caldera, and so is stated in the copyright notice.

see https://www.tuhs.org/Archive/Caldera-license.pdf

Had you run it in simh, you’d get that a pdp-11/45 holds for an amount of around 256Kb max, and a virtual memory space of 64Kb text, and 64Kb data/stack. This probably makes vi completely useless, so you’d finnally end working with ed.

ed has a problem at first. You tend to think you’ll never end the task. But after just two days working with it, you’ll begin getting confident and you’ll have a better idea of your document, will begin to write commands like

/printf(“Hello,/s/”Hello/”Good bye,/

and the like. You say the interrupt key cannot be adjusted. You are right, but a little digging will show you that it is implemented in the kernel, so a short patch to stty(1) command will allow you to solve the problem (As I did on mine, but using ed(1)) Below is the patch. Just run
$ ed <stty.c.patch stty.c

and here is the patch file:

$ diff -e stty.c $HOME/stty.c
229a
ioctl(1, TIOCSETC, &cchars);
.
210a
if (eq("intr")) {
if (**++argv = '^')
cchars.t_intrc = (*argv)[1] & 037;
else
cchars.t_intrc = **argv;
argc–;
}
.
185a
ioctl(1, TIOCGETC, &cchars);
.
178a
struct tchars cchars;
.

after that, you'll have the possibility of setting the interrupt key to whatever you want.

Another thing that surprised me is how to import/export things from that unix. Well, the copy I'm running on simh allows me to write tapes, and I wrote a simple program to convert the tape format of simh (fully documented on the simh documentation, which, by the way, is misteriously written in Microsoft Word format 🙂 ) and this way I was able to produce and export tape files from my unix v7.

The version I'm running is a Dennis Ritchie prepared tape, found in minnie.tuhs.org, under the name unix_v7_rl.dsk, I can provide you a full compliant simh configuration file and the disk images I use (already patched if you wish) As today, they are under Caldera License (and not the Open Group) but the Open Group didn't exist at the date of release of V7 (1978)

My approach was different than yours, I wanted the original code, to see (and to show to anyone interested on it) how all these brave men did it. Vi is very interesting. But some practice on ed will give you a different view of the thing. Editing roff source files is quite easy with ed, as everything is left adjusted in source. Editing C source files is a different thing, mostly if you want your source to be legible after all.

Another thing you need to patch is the date(1) command. It doesn't allow you to enter dates after 1999 (it uses only two digits for the year, and it doesn't convert years after 1999 —nobody expected it to be alive after that date 🙂 ) My first edit was to patch the date command, and I have to admit, I loved these men, that were able in such conditions to write the full unix kernel. You just have to insert a line that says:

if (year < 70) year += 100;

at some point in date.c (you'll find it by yourself) but just give it a try, and try to conserve the source code legibility. I get completely annoyed when I listen at some guys saying that old unix code sucks, because these guys used s (for source) and t (for target) as variable names, or the like. You need some ed(1) training is what I normally answer.

Well, this was my two cents contribution. Now you have for a next part blog. This time write things using ed (after some practice, you get to it) I have also patched getty.c to allow for /etc/issue showing at your console, and LOGIN and USER environment variables defined in login.c Your TERM variable, although has to wait to be initialized in your .profile file, but trying to fit vi(1) in 64Kb+64Kb is just a pain (i will probably do that someday, but for the moment I prefer to export this to my host and use git to maintain source code versions 🙂 For now I'm trying to be able to compile a kernel version adapted to use larger disks (rp) than rl disks (max 10Mb) so I can fit all the system plus sources in a single disk and use rl disks for swapping.

If you need some support, don't hesitate to ask me.

The only thing open about the open group is their willingness to take that 100,000 USD and call anything a Unix.

Remember when some Chinese outfit paid it to have Linux called Unix?

I suppose that was bad wording on my part. I meant it that Bell Labs no longer has the trademark. Trademark misattribution would possibly anger the current trademark holder (The Open Group), and that’s not a legal risk I wanted to carry.

Truth. Bell Labs is the true originator and thus copyright holder of UNIX. The Open Group came much, much later, an ungrateful heir.

That is what I had thought as well. I tried to ask microfocus a while back for a SYSV license, but they only wanted to sell cobol.

The copyright situation with UNIX is complicated and interesting. AT&T no longer owns the UNIX copyrights, and hasn’t for a very long time. Novell owned them for quite a while, following AT&T’s sale of Unix System Laboratories to Novell in 1995. Most recently, Novell was purchased by Micro Focus in 2014. I’m not aware of Novell ever giving up the UNIX copyrights (see: SCO vs. Novell), so I presume the copyrights now belong to Micro Focus.

AFAIK AT&T’s DWB (Documenter’s Workbench) was not part of AT&T UNIX. On the vetusware (no direct link!!! ) there are a sources for the DWB in the Dynix PTX and National Semiconductor 32000 SYSVR2 sources.