I wandered about this a while ago, when VSCode went into Web. Later, after realizing how much download had to take place to have a remote OS ran, again the experience is very hacky with current hardware to software friction. It needs lots of drivers and resources to align experience with an actual OS.
As the author of this post, I have to say that I, too, was surprised by how well it's doing. But people at HN are generally very receptive to Lisp/Smalltalk posts, and lately, at least one Smalltalk-related post has appeared on the front page every week.
I upvoted it because I have fond feelings and memories of Pharo. The warm fuzzies aside, even if someone hasn't used it, it's one of those places that remind us of the delight of programming and computation. Like, "Oh, yeah, I used to feel optimistic and happy about the Computer when I was programming in basic a thousand years ago."
Nice for those trying it from locked down machines. It's a common situation people spending most of their screen time on computers where the user has no admin privileges.
It seems to break when zooming in/out on Firefox (Linux). Zooming in reduces the size of the viewport and makes clicking the mouse happen somewhere else.
So I assume this is the byte code interpreter compiled with Emscriptem? Didn’t Squeak have a similar try it out page?
I would assume there’s no attempt at Jitting or pre-compiling some of the image code, which would be a feat given Smalltalk dynamic nature and webassembly limited execution model.
I was hopping to have something like this to play with. How can you build the code?
I played around with Pharo some time in 2017 or 2018, making a Minesweeper clone in it.
I really _really_ like environments where rapid and immediate iteration is possible, especially for any project with any artistry or exploration in its construction. One thing is that (in some cases) you see intermediate, emergent, or unintended results you'd not have seen otherwise.
Although not a particularly interesting unintended result, a logic problem (iirc) had the Minesweeper expand happen with each action (i.e. propagating cleared cells with each step rather than in an instant).
I've never used Pharo before, but in my firefox it renders up and to the left too much. I'm not sure how the mouse is translated to the canvas but it seems like there is an issue with it (or vice versa and it's the weird canvas scaling). Removing `transform: scale(calc(1 / var(--world-dpr, 1)));` from the canvas css seems to fix it for me though.
I still remember the Byte issue on Smalltalk from 1981 [0]. To high school me it was The Future for sure. It looked so cool. I had no way to try it on the family Apple II+ [1] though.
Trying out various Smalltalk-family systems in recent years showed the advantages of implementing the entire system in one language and running it in one image. There's a refreshing coherence and predictability to everything, and pervasive simplicity instead of fractal complexity.
You can browse the entire codebase in a single window - language/compiler, math operations (except for the base primitives I believe), data structures, graphics, window system, all nicely organized. Methods, even critical ones, are mostly small, straightforward, and cruft-free. Many are just a line or two.
Smalltalk gives the developer immense power. You work directly "in the code"; your changes are live instantly. You can trivially modify the system itself - just go there and start typing. All in the same browser window. You can easily look inside any running program, change its data and even its code, and just keep going. Even now it feels futuristic.
You get powerful graphics and a window system out of the box instead of the usual story where you select OpenGL/Vulkan/GUI/DOM libraries and configure tools like make/rpm/docker/pip/npm. That's probably the most "science fiction-y" aspect by far.
In short, this was the "Jetsons" future that seemed possible back when I was getting started. The future we actually got, of course, is far less elegant and unified, and far more messy and practical. That's probably for the best. But I still love Smalltalk.
Can't believe what I'm seeing with my own eyes that Pharo is this fast, on a browser, on a mobile. I mean I get it that things have advanced and so much more insane things are out there (like running OS X Tiger with Infinite Mac, or SketchUp web, or...) but this still shocks me. Tech is advancing too fast for my monkey brain to pace with.
The live-image real-time programming made possible by Lisp and Smalltalk/Pharo, at least to me, is how programming should be done.
Much as I admire Wirth and his languages, and I understand the uses and necessity of a structured programming, live-programming is the better and more creative approach not least because it feels like writing an essay-- you start and let your ideas take you where they would, instead of forcing yourself to seek your ideas out of your subconscious one by one.
simultsop | 22 hours ago
pjmlp | 20 hours ago
https://developer.microsoft.com/en-us/reactor/events/27557/
simultsop | an hour ago
pjmlp | 21 minutes ago
It is like telling someone that "C went into UNIX" after learning it using Code::Blocks on Windows.
rvz | 22 hours ago
Use with caution.
[OP] xkriva11 | 22 hours ago
DTrejo | 21 hours ago
[OP] xkriva11 | 20 hours ago
echelon | 21 hours ago
Several of my friends love it.
da02 | 21 hours ago
jrootabega | 20 hours ago
melvinroest | 18 hours ago
[1] https://yesplan.be/en
norberthartl | 16 hours ago
lynndotpy | 19 hours ago
jsrcout | 17 hours ago
treyd | 22 hours ago
stronglikedan | 22 hours ago
AdmiralAsshat | 21 hours ago
Qem | 19 hours ago
andsoitis | 10 hours ago
Did a quick check in their manual (https://books.pharo.org/pharo-by-example9/pdf/2022-03-26-ind...) and on page 9 it confirms as much.
popcar2 | 21 hours ago
simonsarris | 21 hours ago
nathell | 20 hours ago
ajasmin | 20 hours ago
So I assume this is the byte code interpreter compiled with Emscriptem? Didn’t Squeak have a similar try it out page?
I would assume there’s no attempt at Jitting or pre-compiling some of the image code, which would be a feat given Smalltalk dynamic nature and webassembly limited execution model.
I was hopping to have something like this to play with. How can you build the code?
fniephaus | 15 hours ago
lynndotpy | 20 hours ago
I really _really_ like environments where rapid and immediate iteration is possible, especially for any project with any artistry or exploration in its construction. One thing is that (in some cases) you see intermediate, emergent, or unintended results you'd not have seen otherwise.
Although not a particularly interesting unintended result, a logic problem (iirc) had the Minesweeper expand happen with each action (i.e. propagating cleared cells with each step rather than in an instant).
Cieric | 19 hours ago
thrance | 19 hours ago
mike_ivanov | 19 hours ago
bmacho | 18 hours ago
mike_ivanov | 17 hours ago
mike_ivanov | 17 hours ago
jsrcout | 17 hours ago
Trying out various Smalltalk-family systems in recent years showed the advantages of implementing the entire system in one language and running it in one image. There's a refreshing coherence and predictability to everything, and pervasive simplicity instead of fractal complexity.
You can browse the entire codebase in a single window - language/compiler, math operations (except for the base primitives I believe), data structures, graphics, window system, all nicely organized. Methods, even critical ones, are mostly small, straightforward, and cruft-free. Many are just a line or two.
Smalltalk gives the developer immense power. You work directly "in the code"; your changes are live instantly. You can trivially modify the system itself - just go there and start typing. All in the same browser window. You can easily look inside any running program, change its data and even its code, and just keep going. Even now it feels futuristic.
You get powerful graphics and a window system out of the box instead of the usual story where you select OpenGL/Vulkan/GUI/DOM libraries and configure tools like make/rpm/docker/pip/npm. That's probably the most "science fiction-y" aspect by far.
In short, this was the "Jetsons" future that seemed possible back when I was getting started. The future we actually got, of course, is far less elegant and unified, and far more messy and practical. That's probably for the best. But I still love Smalltalk.
[0] https://archive.org/details/byte-magazine-1981-08
[1] https://archive.org/details/byte-magazine-1985-05/page/155 (At the time I was sad because we didn't have the extra memory it needed. Now I'm more impressed that it was even possible.)
nater5000 | 17 hours ago
pmkary | 17 hours ago
EmX_3000 | 14 hours ago
jan_m_savage | 9 hours ago
Much as I admire Wirth and his languages, and I understand the uses and necessity of a structured programming, live-programming is the better and more creative approach not least because it feels like writing an essay-- you start and let your ideas take you where they would, instead of forcing yourself to seek your ideas out of your subconscious one by one.