Sorry for being late this week. A plethora of problems, networks that not work, kaput hardware and self-corrupting operating systems seem to have been bugging me this last week. So, a bit late, but here we are:

Professor Andrew S Tanenbaum has had a significant impact on the world of operating systems as we know it today. In order to teach operating systems, he wrote his own clone of Unix, using the most modern design techniques he could think of. (He called it Minix, and that’ll be my OS for next Thursday). This OS in turn was used by a student to start a hobby project. That hobby project became known as Linux.

What he went on to do next is potentially far more interesting and exciting than that.

The main idea of Amoeba will be familiar to you if you run any of the BOINC projects like Seti@Home. The idea is to turn multiple, inexpensive computers into one single transparent system. Imagine it this way – add a hundred thousand Commodore 64s together and you get Deep Thought.

People have been playing around with distributed programming like this for ages – witness the language Occam, for example – but this didn’t really hold enough scope for the Amoeba system. So they designed and wrote their own programming language – Orca (which I’ll talk about later in the week) – designed to make this task easy (well, easier than in Occam anyway).

It’s basically a research OS (which for those of us not researching parallel operating systems essentially means nobody’s ported Firefox to it yet), although the interface is a standard X window interface (X11R6, basically). However it can be used with a wide variety of POSIX based utilities through a library called Ajax (POSIX, if you’re not aware, is the name of the ISO standard for the behaviour of Unix-like operating systems).

The Amoeba site hasn’t been updated in a while, so if you happen to have a few old 486 machines lying around it might be fun to play with. It’s another one on my to-look-at list, so if I make any progress I’ll let you all know…

I recently got my registration code back for ElcomSoft‘s excellent Advanced Disk Catalog (I’ve been using it since Vladimir Katalov himself was in charge of it) and decided to rebuild my catalogues, which I haven’t had around since a hard disk corruption two or three years ago. While doing so, I came across an old PC Plus coverdisk from February 2001, containing Serif Page Plus 4 and QNX.

Well, well, I thought – an entire operating system to play with…

It was obvious from the file structure that they’d just copied it across to a subdirectory on the CD, as the readme referred to the “qnxrtp” directory being a subdirectory directly off the CD. So, a little fiddle with Nero, make the bookdisk image the boot off the CD, use the Image Recorder, and five minutes later I have a bootable ISO image of QNX.

The next question was how much space to use for QNX? Well, all 1GB that I’ve assigned to the virtual disk (the readme said 300MB should be sufficient, so that should be okay) and suddenly… it all stopped. Says it can’t find “/boot/fs”, even though it’s there. But in upper case.

Hmm… bring out Azri Rosborg‘s “Up or Low”, set all to lower case and play with Nero again…

This time it copied five files and told me to remove floppies and CDs and reset. Looking good… until it told me it couldn’t find qnxbase. Hmm. OK, try again, this time in verbose mode. Nothing unexpected except it says if it doesn’t work, press ESC to skip DMA. So let’s try that.

Same messages – “Unable to mount a QNX filesystem… Unable to access packages”. Hmm.  Okay – try it in Virtual Box.

Nope.  Fails even worse.  Won’t even boot off CD.

Oh well. Perhaps I’ll go download the full QNX and ask for a non-commercial licence instead.  In the meantime – FAIL 

The source code for Multics has been released now, “for any purpose and without fee”. The importance of this operating system can’t be underestimated; so many things we take for granted as standard features in modern operating systems were either invented for, or pioneered in, Multics: dynamic linking (DLLs, basically) and online reconfiguration (plug ‘n’ play, but more so: you could even add and remove CPUs while the system was running!) to name but two.

What really set Multics apart though was its security level, which was higher than any system at the time, and is still the only operating system to reach the B2 Orange Book security rating from the National Computer Security Centre (part of the NSA). It was also one of the first – if not the first – OSes to be written in a high level language. PL/I, to be precise (a language notorious for being really hairy).

The significance of the source release is this: Multics, which only ever ran on a specific set of mainframes, may yet come alive again. Possibly the most secure operating system of all time, and certainly one that was beloved by its operators (who call themselves Multicians), may be coming to a PC near you in the next couple of years, but this time as a free, open-source operating system. All it needs is someone to start the ball rolling and start a project on sourceforge… any takers?