UNIX - A History and a Memoir - Brian Kernighan's latest

[PDP-8]

Brian Kernighan - one of the ATT researchers - basically the "voice" of Unix since it began has released yet another great book.  Maybe you've read some of his others over the decades.

UNIX
A History and a Memoir
Brian Kernighan

Even though the history of Unix is well documented on the net, this is not just a simple rehash of all that, nor just pining for the good old days.  Professor Kernighan describes and provides some great detail whether you are just a history fan, or reading between the lines - still see the spark of it all.

Wonder what is was like working at ATT research back then?  He's got some great detail that go beyond recent videos and the like.

So it's not just a dry technical history, but it's like being invited to have a down to earth talk with someone who was there practically from day one.  And how it makes your cellphone, nay even *CORE still relevant.  Good ideas live beyond changes in hardware.

A *book* ? What is that?  It's how we used to do it.  Get in the comfy chair, get some good backlighting on, and spend an enjoyable afternoon reading.  A lost art.

Might be of interest especially to those of us in the *core crowd, which expouses many of the same values.

[hiro]

not a lost art. tinycorelinux is so tiny that now i have more time to read than ever!

[PDP-8]

Me too but I won't be able to fully grok Tinycore until I retire.

There's an opportunity for a good tech writer with say Amazon publishing - although pics usually are just black and white.

Without getting too cornball about it, I find that underneath the hood of ALL the 'Core projects, a simple "elegance" about the underlying infrastructure so fascinating worthy of study.   It isn't just a "me too" application launcher.

I always wonder - what would some of the early luminaries like Brian Kernighan, Doug McIlroy, Rob Pike, Steve Johnson, John Mashey, Steven Bourne and the rest of the crew (most are still around) would think of Tinycore as it relates to today?

I'd think they'd find it fascinating.

[hiro]

they probably don't like that it's linux, but they should certainly be able to appreciate the simplicity compared with other distros.
from what i heard of pike and kernighan they are rather involved in google web services these days, probably trying as hard as possible to use their language golang to make this less painful

[hiro]

also they will probably appreciate our usage of ln -s to bind the libs and binaries of packages into the right places in the root filesystem. they invented this idea (albeit more elegantly) and i'm using tinycorelinux partly bec. this reminded me of how it's done in plan 9.
i.e. there is no PATH in plan9, everything is union bound into /bin.
it's a little bit like ln -s /amd64/bin /bin, just that with ln we need a hack and ln every individual file, to get around that we cannot "union-link" directories.
oh, and yeah, they hated ln. bec. it's not so flexible, for example in this regard (and it's a completely different layer, in the on-disk filesystem!)

[PDP-8]

Most of them just use whatever is convenient - BSD's, MacOs, Linux and whatever turns them on.

My main interest in the history is mostly pre-divestiture of ATT.  The history of pipes everyone knows, but some don't realize the convenience of "everything is a file" - like device files you could merely cat to, without having to drop into a Job Control Language subsystem, answer about 20 questions about the file specifics, and print or save.

cat myfile > /dev/lpr

Mindblowingly convenient.  Suddenly you don't have to be the device-driver with a lot of arcane printer device knowledge as you would with other systems of the day.

My biggest history kick is how McIlroy showed back in '86 or so that most everyone is still using 1960's JCL techniques, when using DD with IF= and OF= pairs.

Nowadays, I just do it the unix-way:

Code: [Select]

dd < input > output
I make far FEWER dd mistakes using the unix notation - it just reads right to me, whereas with IF / OF pairs, I easily get them mixed if I'm not paying attention.

That blew my mind because even today, everyone uses the JCL way of doing it from the 60's.  I thought that was the only way until a few years ago, and just smiled how this could be hidden unless you dug up the history...

I'm weird I guess - all this stuff we take for granted.  Imagine running unix without any pipes and having to use redirection and temporary files to delete all the time?  Or no device drivers - dropping into a subset language of it's own just to manipulate files getting them into and out of filesystems based solely on a manufacturer's spec-sheet - all of them different and YOU have to get it right...

Dunno' all these years and it still seems so innovative to me...

[hiro]

yep, agreed. i knew this much when i first saw your username pop up here

[PDP-8]

Heh, believe it or not, I have never touched a PDP-8!  My first computer was a Vic-20.

One thing missed about pipes was that it wasn't just a cool thing - it was totally practical from a limited-resource unix standpoint on machines of the day with low ram.  We forget that the PDP7 / 11's were the "raspberry pi's" of the day compared to the PDP-10, which management denied them after pulling out of Multics....

So you have a text-file you want to process that is a whopping one-megabyte big to analyze, like the Federalist Papers and feed it into your piddling PDP-11.  Without pipes, you couldn't, since trying to load that data into an editor would max out the ram and crash.

But with pipes handling just a stream going through filters, you could!  And heh, back in those days there was no virtual memory.

Anyhow - I'll stop - I just sit back in amazement at what we have now, but made all the better by knowing what a PITA it was to compute back in those days by researching very early history.  Even though the code is not ATT, the principles still live on and fascinate me to no end.

[Rich]

Hi PDP-8

... So you have a text-file you want to process that is a whopping one-megabyte big to analyze, like the Federalist Papers and feed it into your piddling PDP-11.  Without pipes, you couldn't, since trying to load that data into an editor would max out the ram and crash. ...

That' not a fair statement. Now you are just comparing what a pipe can do to what a broken editor can't do. Back in the early 80s we
used Intel Intellec Series II blue boxes running the ISIS operating system. If you were paging through a large file with the editor (aedit)
the 8" floppy drive would periodically start clanking as more data was requested. This was a machine running an 8080/85 with 64K
of RAM, half of which was used by the operating system.

[PDP-8]

Ah, I didn't know that!

uP development and o/s well prior to the well-known msdos saga always interests me too.  Early stuff, and the people behind it.  And not just Gary Kildall either.

You probably know this already, but a good page on the Intel stuff I browse occasionally for updates (Isis and so forth):

http://www.retrotechnology.com/dri/isis.html

All this stuff was so interesting before it just ultimately became home-computer marketing, joysticks, and game cartridges... and then finally blossomed into unix workstations and the like...  seeing the minis and micros converge was very cool..