I used to work at Cycorp, a holdover from the AI world of the 80s and to this day a classical (CL-like) Lisp shop. They have an old Symbolics box in their entryway, positioned by a couch as a small coffee table :)
I find the whole idea of hardware that's specifically optimized for a totally different programming paradigm than what we're used to just fascinating. It's not hard to imagine why we haven't seen more of it: it's really expensive to iterate on a branch in the computing hardware tree, and you're probably better off just writing a runtime for mainline systems. But still, it's fun to think about.
I am a little surprised though that nobody's written an operating system like this for standard hardware. Still a big task, but an order of magnitude less ambitious than custom hardware and with many of the benefits people would've gotten from a Lisp machine.
Thanks for mentioning Pharo Smalltalk, I wanted to widen the conversation also.
I had a Xerox 1108 Lisp Machine from 1982 to about 1986 (did a successful commercial expert system product and used it for technical demos for marketing).
I think that Pharo is the closest thing today to that good old Lisp Machine feeling/experience. An amazing development environment.
Today, I pay for LispWorks Professional that while a great developer experience, is not like a Lisp Machine. I find it easy to get people to pay me to do Lisp development, but not Smalltalk.
That would be Jazelle, I think handful of Sony-Ericsson phones supported that. Nokia was not a big proponent of Java (they had Symbian to sell), and that was all before Androids time.
As ex-employee I can tell you that Nokia was surely a big proponent of Java, before LWT appeared, the Nokia UI framework was the best extension to J2ME, hardware accelerated on their devices.
They also add J2ME extensions for Symbian capabilities and N95 was the first handset to have real hardware acceleration to J2ME 3D features.
It was a big shock to everyone the communication to move beyond Java and C++ into C# and adopt WP7.
I was told that Jazelle was not popular because it was complicated/expensive to license and slower than JITing. (Basically Jazelle was a hardware interpreter for a subset of Java byte code.)
It was slower than JITing, but the kinds of J2ME feature phones that ran it weren't using JITing runtimes anyway. And it was for sure faster than interpreter loop they otherwise would have.
I used to work with you there and you were interesting for the only person building useful software in the whole org. Soon you will learn to hide the fact you worked there.
For those who don't get the lisp machine. imagine how you can inspect your browser, see the html/js, go to console, run commands modify programs etc. Imagine doing that on your OS, that's the experience of Lisp machine.
I feel that way everytime I use <browser> devtools.. and that's why I think the web is gonna be the only reincarnation of these lisp machines / smalltalk images
Man that looks so cool. What a beautiful, clean user interface and I assume an amazing experience, with the full power of Lisp.
Is there any chance to get a Lisp Machine as an image that we can load in a VM?
All I need is a connection the internet and I would be happy to spend all my spare hobby programming time dialed away in my own small corner of a Lisp Machine. No distractions, only Lisp.
The Symbolics Lisp keyboards are also such amazing inputs devices!(Like the Space Cadet Keyboard https://en.wikipedia.org/wiki/Space-cadet_keyboard)
They had a button or meta key combination for what seems like thousands of characters or functions. I actually netted an old Symbolics 364000 keyboard from my university and I love the design of it, makes me wish we had kept some of that design language in our modern day languages.
They are just not. The Honeywell switches feel like crap IMO. There are no cursor keys. A 104 key Unicomp has better switches and better ergonomics, and just as many function keys - the only thing missing is a 5th (Symbol) modifier key. All you need is the proper key mapping. A 109 key Japanese keyboard is going to have more modifier keys available, and you can get them in your choice of switches.
Please don't part out keyboards from old computer systems so you can use them with your Macbook. Give them to people who are actually working on collecting and preserving the old systems.
> Please don't part out keyboards from old computer systems so you can use them with your Macbook. Give them to people who are actually working on collecting and preserving the old systems.
...Or do with them as you please, so that they may continue pushing the world forward instead of uselessly sitting around in a museum corner collecting dust.
100% agree, we have full documentation of the board schematics and have photos from full teardowns, along with examples already stored in museums. Storing becomes hoarding after a point.
I am not talking about hypotheticals or museums, I am talking about actual running systems. I know the most active restorer of Symbolics systems in the US. His biggest problem is finding enough consoles and keyboards.
Why don't you try to actually build a reproduction Symbolics keyboard (I did), and see how much time and money it is going to cost you.
They aren't tactile, but they weren't meant to be, they were utilitarian. Under normal conditions hall-effect switches could last up to 30 billion operations (From Honeywell's own documentation https://sensing.honeywell.com/hallbook.pdf). They are designed to last and be used for the entire life of a machine, through all its upgrades. The lack of cursor keys was normal for the time, just look at the IBM Model F.
As for Unicomp, they are based off of the IBM Model M (Which I have a few of and love the buckling spring mechanism) but they were designed to have feedback (auditory and tactile) built in. For sure a good design, but that doesn't make it the only good design.
>Please don't part out keyboards from old computer systems so you can use them with your Macbook. Give them to people who are actually working on collecting and preserving the old systems.
I never said I did or would? Hall effect sensors are still available if someone wanted some they have much cheaper and easier options than tracking down an old keyboard. Also who ever implied that I'm not collecting and preserving old systems?
There are 30(!) comments on HN linking to this video, including discussions of Bret Victor's "Learnable programming", using Lisp in production from last year, comparisons with Smalltalk, a thread with examples of beautiful software, and more: https://ampie.app/url-context?url=https://www.youtube.com/wa...
Well, their arguments[1] are quite compelling in general, rest of the software[2][3] seems to be highly specialized and specific, I'm guessing there is a lot of maintenance that has to be done still.
- [3] http://www.symbolics-dks.com/PDease-1.htm (general purpose software that uses Finite Element Analysis to obtain numerical solutions to a large class of partial differential equations)
I believe those instructions do not work anymore. There's a project on github that will build an appropriately old ubuntu vm that runs open genera pretty well provided you have the source. You can also find the source tarball without too much hassle on some torrent sites.
I find the whole idea of hardware that's specifically optimized for a totally different programming paradigm than what we're used to just fascinating. It's not hard to imagine why we haven't seen more of it: it's really expensive to iterate on a branch in the computing hardware tree, and you're probably better off just writing a runtime for mainline systems. But still, it's fun to think about.
I am a little surprised though that nobody's written an operating system like this for standard hardware. Still a big task, but an order of magnitude less ambitious than custom hardware and with many of the benefits people would've gotten from a Lisp machine.