Wow, this is excellent. The Oberon and Bluebottle OS materials have always been quite scattered, so someone putting them in a central index is quite convenient.
For those unaware, Oberon's main qualities are the fact that it's a full operating system written in a garbage collected Pascal-like language (actually made by the same person who initially wrote Pascal) which uses said language's module system to provide reusable/chainable interfaces throughout the whole OS, support for orthogonal persistence and most notably, its highly unconventional user interface that bridges the power of the CLI and the GUI together in this vaguely hypertext-like workspace where you can dynamically live program the UI itself through on-screen text that can serve as an entry point or continuation to perform all sorts of computations, things you'd normally write hacky scripts for. Closest analogue is Xerox's Cedar.
You should consider trying it and stealing a few ideas from it for the greater good.
> on-screen text that can serve as an entry point or continuation to perform all sorts of computations, things you'd normally write hacky scripts for.
The Acme editor famously works this way. So, in a rudimentary way, does the Emacs scratch buffer. It's a weird amalgam of text editor and REPL: you evaluate snippets of text and get the output in-place, and then edit the text to get what you want next. This live-text-as-code way of working is not how I'm used to interacting with a programming system and always slightly rewires my brain. But even in that simple form, you can grasp in it the beginnings of a whole computational paradigm, all the way up to UI. (Computational models don't usually imply a UI—this is an exception, as are spreadsheets.) Oberon must be the most systematic realization of this. I'd like to try it.
Which is unsurprising, because Rob Pike was openly influenced by Oberon in the design of both Acme and the original Plan 9 windowing system 8½ and its current successor rio.
Oberon exploits it further, though. Plan 9 does use chording, programmable text and plumbing but only insofar as it complements the synthetic file system interface, whereas Oberon is more thoroughly object-oriented and can have the text serve as a pointer to various OS subsystems, creating a sort of graphical continuation-passing style, as I alluded.
The acme editor was new to me so thanks for that. The comment about spreadsheets reminded me of an interview on the Thoughtbot Podcast with Chris Granger about a new IDE being developed called Eve. [podcast link](http://giantrobots.fm/111) If I understand correctly, the idea is that its like excel, but you use it to build domain editors, but it revolves heavily around constraint solvers? I don't know, but the podcast episode is awesome. Bottom line is that I'm giddy to use it [link to the announcement post](http://www.chris-granger.com/2014/10/01/beyond-light-table/)
> highly unconventional user interface that bridges the power of the CLI and the GUI together in this vaguely hypertext-like workspace where you can dynamically live program the UI itself through on-screen text that can serve as an entry point or continuation to perform all sorts of computations, things you'd normally write hacky scripts for. Closest analogue is Xerox's Cedar.
Sounds a lot like Smalltalk's print-it/do-it? I suppose the idea orginated with Oberon?
The only thing I didn't understand is the part about garbage collection. What does garbage collection mean and what is the benefit in this implementation?
Programming language with automatic memory management. Look up "garbage collection" for all the details in the world.
The reason I brought it up is because it's a common trope that you can't do low-level programming with GC, and systems like Oberon are direct evidence to the contrary.
Exactly. I see the question on Stack Overflow all the time and just keep dropping a whole list of counterexamples. Too many myths in IT that hold mainstream back.
Hmmm, if you don't know what garbage collection is (and I don't mean this as sneer) how are you sure that you understood everything else said?
In the sense that garbage collection is a pretty basic CS term, and the summary contained lots of CS terms and implied knowledge to understand it (modules, interfaces, persistence, etc).
Actually, as a self-taught developer who started with Python and never had to learn about memory management, "garbage collection" was a concept I only became familiar with after a few years of reading gradually more & more material, usually from HN. The other stuff in the summary above would have either made sense or been "guessable" to 3-years-ago-me.
Garbage collection was never really anything I recall being discussed in my Cmpt Science curriculum, back in 1990-1993, though there was lots of interesting theory about memory allocation/deallocation/defragmentation. It wasn't until I ran into Java around 2000 or so that people started talking about Garbage Collection.
I'm not saying that it didn't exist, it was just less of a priority 25 years ago than 15 years ago.
Everyone has different experiences and interests in college, but
I recall Garbage Collection as an important subject decades before 2000.
I studied CS in the 1970s (and 1980s) and garbage collection was already an
important subject. It was a part of the ACM recommended CS curiculum
as far back as 1968. Google Scholar reveals thousands of results for
"garbage collection algorithm" before 1990. The subject was studied by
great computer scientists (Lamport, Dijkstra, and Liskov are all Turing
award winners) and the results were fun and interesting:
1989
AW Appel: "Simple generational garbage collection and fast allocation"
1988
JF Bartlett: "Compacting garbage collection with ambiguous roots"
J Crammond: "A garbage collection algorithm for shared memory parallel
processors"
1987
AW Appel: "Garbage collection can be faster than stack allocation"
DI Bevan: "Distributed garbage collection using reference counting"
1986
B Liskov, R Ladin: "Highly available distributed services and fault-tolerant
distributed garbage collection"
1985
J Hughes: "A distributed garbage collection algorithm"
1983
H Lieberman, C Hewitt: "A real-time garbage collector based on the lifetimes
of objects"
1982
RJM Hughes: "A semi‐incremental garbage collection algorithm"
1981
J Cohen: "Garbage collection of linked data structures"
1977
HC Baker Jr, C Hewitt: "The incremental garbage collection of processes"
1976
Douglas W. Clark: "An efficient list-moving algorithm using constant workspace"
PL Wadler: "Analysis of an algorithm for real time garbage collection"
1975
Guy L. Steele, Jr.:"Multiprocessing compactifying garbage collection", CACM
Edsger W. Dijkstra, Leslie Lamport, et al.: "On-the-fly garbage collection:
an exercise in cooperation"
1974
Gary Lindstrom: "Copying list structures using bounded workspace"
1973
Edward M. Reingold: "A nonrecursive list moving algorithm"
1972
H. D. Baecker: "Garbage collection for virtual memory computer systems"
Ben Wegbreit: "A space-efficient list structure tracing algorithm"
1969
Robert R. Fenichel, Jerome C. Yochelson.: "A LISP garbage-collector for
virtual-memory computer systems"
Joseph Weizenbaum: "Recovery of reentrant list structures in SLIP"
1967
H. Schorr, W. M. Waite: "An efficient machine-independent procedure for
garbage collection in various list structures"
Peter J. Denning: "The working set model for program behavior"
1963
Daniel J. Edwards: "Secondary Storage in LISP"
The Communications of the ACM, during this period, was the most read
CS Journal. Articles appearing there were intended to be of interest
to the widest population of computer scientists and 19 of them were on
garbage collection.
LISP, APL, LOGO, ML, Prolog, CLU, Scheme, Cedar, Smalltalk, Icon, SML,
Mathematica, J, Haskell, Python, Dylan, Self, Lua and Javascript are
all important, garbage collected programming languages that came
before the late 1990's.
A reminder of how curious it is that PARC had all of Kay's, Smith's and Wirth's handiwork under their belt, all the rudiments of today's PC with mouse-controlled icon WYSIWYG GUIs back in the late 1970s, and never had the imagination to go anywhere viable with it (but give it away as dramatized in 'Pirates of Silicon Valley'). Same could be said of PalmPilot and the smart phone.
I do Xerox PARC archeology, always looking for documentation from those days.
This was triggered by having been a Smalltalk and Oberon user on their glory days.
It made me realise how much the IT world lost by having UNIX clones going mainstream instead of these systems. Specially the culture that still insist in using their computers as if they had a PDP-11.
Sounds fantastic, but I'm sure it's not. If you made billions on a product such as did Xerox you'd be keen to fully investigate a technology that threatens to supplant your bread and butter product, fear will drive you to do this. Once the product has been developed you will look at it and say "Product X has a return of y, while there are zero customers for this newfangled product", this is more than enough for the board to reject the new idea, many of them think the new technology is just a passing fad anyway.
Good summary. I'll add that Wirth's focus on simplicity makes it very easy to port to new hardware, too. Basically a compiler backend and some low-level stuff on bottom. Usually 1-2 students at ETH would do it over several months for each new architecture.
It is specifically Oberon-07[0], a 2007 revision of the original Oberon from 1986, which is still being worked on by Niklaus Wirth with updates as recently as this year.
Component Pascal is a superset of Oberon-2, itself an extension of the original Oberon that tacked on limited OOP support and a handful of other tricks. Component Pascal was not a Wirth-designed set of extensions, however Blackbox Component Builder is quite the nice piece of kit in its own right.
Unfortunately the availability of modern implementations for the Wirthian languages leaves something to be desired. The only implementations that cover both 64-bit platforms and any semblance of cross-platform support is the Vishap Oberon Compiler[1], Gardens Point Component Pascal[2], GNU Modula-2[3] and m2c[4].
There's also an Oberon-07 to JavaScript implementation[5].
At this point, the very small Wirthian languages community has been mostly focused on Oberon-07. It's worth noting there are a few Modula-2 fans about still poking at things, such as the Modula-2 R10[6] effort.
> Unfortunately the availability of modern implementations for the Wirthian languages leaves something to be desired.
There is the Delphi-compatible, GPL-licenced Free Pascal¹). GNU Pascal²) is also still around but might or might not meet your definition of 'modern implementation'.
Oberon. Now that's a name I haven't heard in a long time. Once upon a time there was a great crossroad in the early 90's (when I was a kid and learning programming - keep in mind that perspective).
BASIC was on its way out - we all knew it to some extent due to home computers renaissance, but it was evident it didn't have any staying power. So there were all these wonderful machines with different architectures, OS' and programming languages out there, with no clear winner (it depended on what you wanted to do). So, basically there were two camps in the end, regarding programming languages. Pascal and C. Pascal had that notion that it was, too, on its way out, but was really useful and Oberon was around the corner so it was worth the wait to stick with it. On the other hand C++ was entering the arena full force, because machines were getting faster (it had a stigma of being slow). I went with C, because I was getting into SGI/IRIX and CG (and later on abandoned programming as a full time choice), and some of my friends went with Pascal. Oberon was floating in the air for some years, but nothing happened. This was always a mystery to me. Eventually, Pascal guys moved to Delphi and my circle of C guys either stayed with C (like I did) or bought OO Kool-Aid and went with C++ (of which some later on went to Java). Pascal (later on Delphi) guys developed sort of a cult. It never was clear what happened to Oberon, and (to me) to this day it's a mystery. Funny enough, I can now retrospectively see that programming languages we chose (and stuck with) was heavily influenced by machines/OS' we used. Pascal guys were mostly PC or Atari guys, and C were mostly those with access to *NIX and Amigas.
Also most of my friends with Amigas cared only about Assembly.
Of course. I tried to be as brief as possible, not to bother people. Even C had a stigma of being slow back then. Assembly was the language for anything performant, but everybody was aware that in the following years computers will get fast enough (computers, not compilers hah) not to care.
That is why I always smile when people talk about how modern languages are so slow and C is the king of speed, if only they used those old C compilers back then...
> Also most of my friends with Amigas cared only about Assembly.
The irony in that is that the Amiga had a lot of decent implementations of Wirths languages. Both Pascal (including the moderately comercially successful HiSoft Pascal), Modula-II and Oberon.
> Oberon was floating in the air for some years, but nothing happened. This was always a mystery to me.
Oberon was an operating system as well as a language, and Oberon was an operating system without processes. Game over.
I mean, there are other cons, like the UI. But in a world where Unix was a done deal, NT and Plan 9 were being spun up, BSD was breaking free and users were happy enough with their Unix, MacOS and Windows machines, Oberon took simplicity too far in a number of ways. (Pascal and Modula-2 were a better bet)
In a weird coincidence I saw that someone is selling an Oberon system today and advertising its single-process nature as a feature. Some kind of networked realtime finance application or something, pretty niche, and who knows if they'll make any money with it.
Oberon had so many good ideas. It's still worth studying.
The problem in the 90's: provide executable content across the net for browsers.
Java was supposed to provide the portable universal binary code you could load and execute everywhere, except that it did not have the necessary features and was too complicated. Then came Javascript but it was broken mess for long time and needs binary format.
There was Juice back in 1997 http://www.modulaware.com/mdlt69.htm. Ligthing fast single pass compiler that works with AST and gives constant-time type and well-formedness checking portably over the net.
If WebAssembly is ready in 2017, we can finally have the portable binary with the same set of features as Juice 20 years later. Instead of Oberon system, we have browser in the client and node.js in the server throwing WebAssembly around.
To tie Juice into the modern world of Javascript and Web-assembly:
This comes out of Michael Franz' PhD thesis on Semantic Dictionary Encoding. Franz did his PhD under Wirth, and is now a professor at UC Irvine where he was the thesis advisor for Andreas Gal.
Andreas Gal was instrumental in getting tracing JS JITs of the ground, and until recently served as CTO at Mozilla.
The problem with Java binaries on the client, was that the Java applet didn't have direct access to the DOM bindings.
Instead, the applet rendered it's own GUI via the plugin, and the look and behavior was inconsistent with the rest of the browser experience.
A similar story with activex and flash.
It's a mystery why it has taken so long to get to the point where we can target the browser with a statically-checked higher-level language that has API access to the native browser event loop and dom components the same way that Javascript does it.
> It's a mystery why it has taken so long to get to the point where we can target the browser with a statically-checked higher-level language that has API access to the native browser event loop and dom components the same way that Javascript does it.
We are already there. Compile to JavaScript (or, heck, WebAssembly if that ever becomes feature-complete enough).
It was a great project. Still faster than Javascript if run through an optimizing compiler. Combine that with today's best security schemes for browsers and you have a lot of performance for a better level of safety. Not going to happen, though. Not enough overhead & buzzword compliance to make it into a web stack. ;)
I had the chance to take a Oberon related lecture and its simplicity is really nice. Some things found in Oberon, like tiling window management, I use everyday. The book describes the whole system including the RISC CPU design.
As part of the Niklaus Wirth Birthday Symposium to celebrate his 80th birthday he also gave a talk titled "Reviving a computer system of 25 years ago" (Abstract[0], Slides[1], Video[2]).
There was also a demonstration on the original hardware, the system really seemed ahead of its time.
Project Oberon also inspired me to write a text editor[3] using a piece table data structure as described in Chapter 5.
Project Oberon is a design for a complete computer system. Its simplicity and clarity enables a single person to know and implement the entire system, while still providing enough power to make it useful and usable in a production environment.
Yes! We need so much more of this.
If a system is to serve the creative spirit, it must be entirely comprehensible to a single individual.
Lisp Machines. I am reading the book, 'The Architecture of Symbolic Computers' written in 1990 by Peter M. Kogge. Amazing book. It's really teaching me a ton of computer science at a level in one shot that no other book has. People fail to realize at times, that Linux was a choice in the road, and people just stuck with it. There could have been so many other ways to go: smalltalk machines, Lisp machines, Minix vs Linux, etc... Love this stuff.
This isn't the same, but if you're interested in looking at a compiler and OS mostly from scratch based on C check this out. It's pretty amazing. https://github.com/rswier/swieros
It comes from the author of the c4 compiler. (Interpreter?)
The work started when Wirth saw the same Xerox setup that Apple did. Wirth couldn't buy one so he just built his own lol. The first system Wirth and Jurg Gutknecht designed was Lilith [1]. They improved Pascal to make modular software in the form of Modula-2. They made a P-code-like assembler language called M-code to make compilation easier and raise assembler abstraction slightly. They then wrote most of the system in Modula-2. They had in about 2 years a computer, OS, compiler, and some apps. ETH used these day-to-day and they later morphed into Oberon system.
The brilliance of Wirth was keeping things simple. I think he overdid it but it served him well in many ways. He also kept things consistent where possible. Just having a simple language, compiler, libraries, and consistent + simple bytecode target would be better than what I've dealt with coding. Each iteration, he tries to improve the language and platform with lessons he learned from the first. He also ensures the lowest common denominator is easy to port, compile efficiently, and produce efficient code for.
A lot of inspiration. He's recently put Oberon on a custom, simple processor running on an FPGA. The latest incarnation of the system is A2 Bluebottle [2] with downloads here [3].
I was going to try Oberon Native again (for the first time since 1996) after getting 9Front to work VMWare Fusion but couldn't find the installer images. I went to the following site:
I don't know why but this kind of project brings me back to my Commodore 64 days. I remember seeing an announcement for a new version of the C64 but it would not make sense unless it could capture the spirit of that little yet powerful 16-Bit device that you can learn so much from and also play amazing games on.
Booting straight to Basic as the CLI inspired many of us to just try and write our own little "software". It must have been the beginning of many CS careers.
This type of inspiring device can only come in a mobile form these days and none of the current OSes are as beginner friendly and straightforward as Basic on the C64 was.
There might be a need for a tablet with an Oberon like system that could attract the teens of today as a customizable and easily programmable device.
An interesting thing about Oberon is that the most recent revision removed some basic elements. Like multiple return statements… Wirth really likes his minimalism.
I am new to all this. I was thinking to get started with The Elements of Computing Systems by Nisan/Schocken. Is Project Oberon something entirely different? If not, how do they compare?
Oberon is an already-built system designed by Niklaus Wirth, one of the giants of the field. The goal of Elements of Computing Systems is to have you build a more limited but still functional system yourself, so get busy!
That's a somewhat self-inflicted wound. Your browser is not sending a user-agent, and the site fails to follow RFC2616 Section 14.43 (sending a UA is _optional_).
For those unaware, Oberon's main qualities are the fact that it's a full operating system written in a garbage collected Pascal-like language (actually made by the same person who initially wrote Pascal) which uses said language's module system to provide reusable/chainable interfaces throughout the whole OS, support for orthogonal persistence and most notably, its highly unconventional user interface that bridges the power of the CLI and the GUI together in this vaguely hypertext-like workspace where you can dynamically live program the UI itself through on-screen text that can serve as an entry point or continuation to perform all sorts of computations, things you'd normally write hacky scripts for. Closest analogue is Xerox's Cedar.
You should consider trying it and stealing a few ideas from it for the greater good.