Hacker News new | ask | show | jobs
by DennisP 1887 days ago
Dang. I used to look forward to the 20KLOC OS/GUI/IDE. I guess they never really achieved that.

Maybe it was just time for Alan Kay to retire.

5 comments

They actually did... to a degree. The last artifact I'm aware of was called Frank[1] and incorporated a number of designs, concepts and DSLs the group presented and wrote papers[2] about over the years the project was active. However, I believe Alan was disappointed that he never achieved the 'from the ground up' deliverable that he really wanted as Frank relied on Squeak as its VM and OS. I think what he really had in mind was to have it at least generate its own low level machine code and possibly synthesize its own logic, where necessary, using FPGAs. (I may be imagining the FPGA part but seem to recall him talking about that at one point)

So while they never achieved the incredibly ambitious goal of everything in 20kloc, or doing it exactly how he wanted to, I think they showed it could be done (at least within that order of magnitude.) The amazing part is the incredible economy of some of the major subsystems they built.[3] Finally, they did assemble it all into a working demo. Unfortunately, without further funding that's where it ended.

[1] IIRC, it was because it was a bit of a Frankenstein solution: parts in Smalltalk, some serious hacks to the Smalltalk image, a couple of VM plugins in C for performance, parts in JavaScript, parts in various DSLs etc.

[2] All of which are publicly available including the code. While it's a bit high level and some of the concepts weren't fully fleshed out, I have managed to get some major pieces of it working with my own code so can attest to the fact that it does work and is useful as a starting point at least.

[3] The first time I really started looking under the covers of OMeta I kept wondering 'where the hell is the code?' After a while I realized that it was all there: it's one of the more amazing uses of recursion and (effectively) self-modifying code I've seen.

Oh that's more than I realized. Is the demo available somewhere?

If they were making good progress and just ran out of funding, that's pretty sad.

This group blew up very suddenly, leaving a lot of unfinished work. They didn't even properly release the work they had done on GitHub or someplace similar. It's just not like Alan Kay to do that.

https://awarth.blogspot.com/

Note that Warth's name is missing from the 2012 STEPS report.

Piumarta went to Japan.

There's something we don't know.

For a bit of follow-on, there is Alex Warth's homepage:

http://www.tinlizzie.org/~awarth/

I had a brief look at Alan Kay's most recent comments here - but couldn't find much about VPRI in particular - I was hoping maybe there'd been some post mortem that I'd missed - but doesn't appear so:

https://news.ycombinator.com/threads?id=alankay

There was a mail list for the project, and no answer ever appeared there, either, despite requests for updates. Very strange.
Edit: chubot found Amelang's github https://news.ycombinator.com/item?id=26927698

Going up a level, there is Gezira and Nile. The latter includes Maru with changes through Dec 2012. Piumarta kept committing changes to Maru through Nov 2013. piumarta.net is, sadly, unresponsive today.

So, it appears that Dan Amelang was able to capture the core of the STEPS project at github. I don't see anything about the UI or Frank, though.

> piumarta.net is, sadly, unresponsive today

Did you want piumarta.com, which somebody else linked above? It's got pages for COLA and MARU

You are right. That's my mistake!
HARC also seemed to end suddenly. I was never able to get any details on why.
Vague memory is funding dried up. Vi Hart has some storytelling touching on CDG / HARC history: https://theartofresearch.org/a-history/
yeah OMeta - interesting idea thanks for noting this, I was unaware (busy with other things lately)
I am pretty sure I saw the demo running in my browser for this, the link was at the end of one of the talks I think. Will post here if I find it
Is it true that they never achieved it? I thought they basically did. You can see the system in action in various talks by Alan Kay.
I would like to see it too, and I even referred to it in a comment a few days ago. Did they release the code or just describe it in papers and talks?

A cursory look at the website and Googling says the latter, which is a shame. The research would be more impactful with source code. There's no reason not to release the source, and if you fail to do so, you can't really complain if practitioners don't pick up those techniques :)

I think OMeta was related and that is released, but IMO it's not that practical.

-----

Thread from 5 years ago that is not that flattering of their work: https://news.ycombinator.com/item?id=11686325

Oh it actually points to some more source: https://github.com/damelang/nile

Although I think it's easy to pick on different parts; there's definitely value to having a holistic system and design. But it would be better if we could see all of it instead of just the parts

They did basically achieve it. There's one missing component though: the kernel. One big reason kernels have to be gigantic nowadays is the stupidly high diversity of hardware interfaces. Every USB device, every graphics card, network card, printer… has its own peculiar way of talking to the rest of the computer, and that shorcoming has to be compensated by bolting a driver on top. For each OS.

See Casey Muratory on the 30 million lines problem: https://caseymuratori.com/blog_0031

Solve that, and then ideas from VPRI can shine.

In the meantime, seems like a VPRI system that just runs on a standard virtual machine would be pretty nifty.