2024-08-25 00:01:26 <unjust      > i've used termux-x11, but x11 forwarding is a poor choice compared to vnc in my opinion
2024-08-25 00:02:06 <xentrac     > but we weren't talking about forwarding anything
2024-08-25 00:02:25 <xentrac     > just running an X client like xeyes talking to an X server which controls the display
2024-08-25 00:03:23 <unjust      > got it. the only time i've used termux-x11 is to have the X server on my phone and the client running on my desktop
2024-08-25 00:04:14 <xentrac     > yeah, VNC is much better at that generally
2024-08-25 00:06:01 <xentrac     > this story is excellent: https://philosophy.williams.edu/files/Egan-Learning-to-Be-Me.pdf
2024-08-25 00:19:53 <MrMobius    > xentrac: no, berrylium is a cat 1 carcinogen
2024-08-25 00:21:59 <xentrac     > from what I said you would expect it to be, wouldn't you?
2024-08-25 00:22:20 <xentrac     > Category 1 doesn't say anything about how common the resulting cancer is, just how good the evidence of causality is
2024-08-25 00:25:39 <unjust      > xentrac: well i just tried using termux-x11 in a "normal" way by invoking local X clients. so i installed xterm and fluxbox, which are basically the core of my GUI usage apart from firefox - which is actually what i had tried x11 forwarding for in the first place. so far, it's A LOT more snappy than what the vnc connection to my laptop is.
2024-08-25 00:26:32 <unjust      > i thought i'd have to install xvkbd or a similar onscreen keyboard, but termux-x11 has that slide-to-the-left-textbox that you mentioned your version of AVNC has
2024-08-25 00:27:31 <unjust      > there're also some mouse buttons shown on screen "L", "M", "R" - but i'll get rid of those when i find the right config setting
2024-08-25 00:27:58 <MrMobius    > xentrac: you said it's basically not carcinogenic so no you would not expect it to be a carcinogen
2024-08-25 00:28:40 <xentrac     > what I actually said was "beryllium basically isn't carcinogenic; it causes berylliosis, which is horrible and incurable but not really a cancer.  of course any persistent injury will cause some amount of cancer but that isn't really the thing to worry about with beryllium"
2024-08-25 00:29:20 <xentrac     > persistent injuries include berylliosis
2024-08-25 00:29:30 <xentrac     > ethanol is also a group 1 carcinogen
2024-08-25 00:29:48 <xentrac     > unjust: does termux-x11 support multitouch?  that's one of the major flaws of the VNC protocol I think
2024-08-25 00:31:05 <unjust      > i'm not sure, - but i don't think it does
2024-08-25 00:31:23 <xentrac     > hmm
2024-08-25 00:31:27 <unjust      > as the touchscreen input modes are: "Trackpad" , "Simulated touchscreen" and "Direct touch"
2024-08-25 00:31:35 <xentrac     > sounds like AVNC
2024-08-25 00:31:41 <unjust      > i haven't tried "Direct touch" yet
2024-08-25 00:32:01 <xentrac     > I guess my multitouch experiments will have to stay in the browser, which is annoying
2024-08-25 00:32:04 <unjust      > but i wouldn't use multitouch gestures if they were supported
2024-08-25 00:32:18 <unjust      > which is because, i don't use them already :)
2024-08-25 00:32:44 <xentrac     > multitouch potentially offers very precise high-bandwidth input; watch any streamer with a handcam playing FPS games on a cellphone
2024-08-25 00:33:16 <xentrac     > but nothing but FPS games and Minetest and Minecraft takes advantage of that
2024-08-25 00:33:40 <xentrac     > using a multitouch screen to emulate a keyboard is like clubbing someone to death with a loaded Uzi
2024-08-25 00:33:48 <unjust      > i play neither of those things
2024-08-25 00:34:05 <xentrac     > you can look at recorded streams on YouTube
2024-08-25 00:34:52 <unjust      > what do your multitouch experiments involve?
2024-08-25 00:34:56 <xentrac     > I've played Minetest and Shattered Pixel Dungeon on my cellphone
2024-08-25 00:35:29 <xentrac     > well, http://canonical.org/~kragen/sw/81hacks/progeome/ is one attempt
2024-08-25 00:35:43 <xentrac     > it's unfamiliar and therefore unintuitive
2024-08-25 00:37:57 <xentrac     > but the idea of this experiment is that you can select verbs (the big blue buttons) with one finger, and while holding it, select objects and provide continuous input with another
2024-08-25 00:38:34 <unjust      > well multitouch at least works on my phone via firefox, i can see the difference in colours depending on how many fingers are on the screen
2024-08-25 00:38:56 <xentrac     > the difference in colors depends on what action is active
2024-08-25 00:39:10 <xentrac     > this supports four actions: delete, move, create line, and create point
2024-08-25 00:39:23 <xentrac     > plus selecting an existing point, which isn't an action
2024-08-25 00:39:37 <xentrac     > but is a necessary prerequisite to creating lines
2024-08-25 00:41:16 <unjust      > ok i get it now
2024-08-25 00:41:40 <xentrac     > I'm not sure it really succeeds at being usable, but I think it at least points in the direction of some really interesting possible user interfaces
2024-08-25 00:41:57 <unjust      > it wasn't immediately apparent what was going on
2024-08-25 00:42:02 <xentrac     > not at all
2024-08-25 00:42:10 <xentrac     > nor is it apparent what you can do
2024-08-25 00:42:52 <unjust      > how viable would this be for some sort of onscreen chorded input?
2024-08-25 00:43:07 <xentrac     > in a sense it is onscreen chorded input, isn't it?
2024-08-25 00:43:15 <unjust      > yes it is
2024-08-25 00:43:25 <xentrac     > but presumably you mean a different sense
2024-08-25 00:43:27 <unjust      > but i mean in a broader sense of acting as a keyboard
2024-08-25 00:44:24 <xentrac     > I don't know.  it's probably possible to do an onscreen chording keyboard that's faster than existing keyboards, but I don't think Progeome represents a step in that direciton
2024-08-25 00:49:20 <unjust      > apart from the fact that i prefer pointing devices and physical buttons - i think onscreen keyboards that try to resemble physical keyboards just waste space at the sort of resolutions you find on 4" - 6" displays
2024-08-25 01:02:59 <xentrac     > well, one of the things you learn from watching recorded video game streams is that onscreen buttons don't have to occupy space
2024-08-25 01:03:54 <unjust      > what do you recommend watching to see a good example of that?
2024-08-25 01:07:17 <unjust      > i'm definitely willing to entertain the idea that i have defective preferences and i might be oblivious to some benefits of touchscreen interfaces that i haven't encountered yet. i'm not really convinved keyboards and trackballs (nor mice and trackpads) are ideal in any case.
2024-08-25 01:14:58 <zelgomer    > i haven't been following the conversation but i'll drop in to say i hate touchscreens with a burning passion
2024-08-25 01:18:39 <zelgomer    > just on the cusp of the smartphone explosion, there was a blackberry with a touchscreen that you would click. that's an idea i wish had continued. other than that, i hate that i have to cover what i want to click with my hand, and there's no way to feel what i'm clicking before i click it. so they require way too much attention to use without misclicks, and even then i still get frequent misclicks.
2024-08-25 01:19:30 <MrMobius    > ya this is why I still haven't given up my graphing calculators
2024-08-25 01:19:54 <MrMobius    > some have awesome keypads and you can program very quickly on them
2024-08-25 01:28:19 <xentrac     > unjust: unfortunately I haven't saved any streamer URLs, I'll let you know next time I see one.  I've seen some others; Bret Victor's multitouch animation program prototype is https://www.youtube.com/watch?v=CfZzCbIKofI and is 2'13", and Haijun Xia has done some other really nice prototypes like https://www.youtube.com/watch?v=WgzdFoXLX9c ("object-oriented drawing"), and
2024-08-25 01:28:23 <xentrac     > https://www.youtube.com/watch?v=FuMcNskrhis ("collection objects")
2024-08-25 01:28:53 <xentrac     > Bret Victor's "Stop Drawing Dead Fish" talk also has some interesting prototypes
2024-08-25 01:29:03 <unjust      > thanks!
2024-08-25 01:29:10 <xentrac     > I don't think it's that you have defective preferences
2024-08-25 01:30:02 <xentrac     > but I think that touchscreens emulating keyboards and mice are necessarily a lot worse than touchscreens using UI idioms optimized for touchscreen interaction
2024-08-25 01:31:11 <xentrac     > I don't know how good touchscreens can get but they can clearly be enormously better than when they're emulating keyboards and mice
2024-08-25 01:31:33 <xentrac     > oh, another appealing touchscreen-based UI is the Reactable, which uses tangible objects on the touchscreen
2024-08-25 01:33:01 <neauoire    > it has been a while since I heard about the reactable
2024-08-25 01:33:30 <xentrac     > for certain tasks I think it's clear that even the best keyboard interfaces are *much worse* than multitouch touchscreens; simultaneous movement, rotation, and zooming is one example.  So we should probably expect that multitouch screens will be *much worse* than keyboards at some things, too, no matter how good the multitouch interface is
2024-08-25 01:34:09 <xentrac     > hmm, apparently https://www.youtube.com/watch?v=CfZzCbIKofI has been taken private since I bookmarked it, sorry
2024-08-25 01:34:21 <unjust      > i don't think i've been exposed to any user intefaces that take advance of those optimizations. so i think i'm unnecessarily closed minded about touchscreen interfaces because i've had years of exposure to user interfaces that didn't make the most of the input technology available (including ones i've created for use via touchscreens that actually did feature multitouch-capable controllers with drivers
2024-08-25 01:34:27 <unjust      > that could have taken advantage of that)
2024-08-25 01:34:33 <unjust      > s/advance/advantage/
2024-08-25 01:34:40 <xentrac     > it's an excerpt from Inventing on Principle
2024-08-25 01:36:14 <xentrac     > specifically it's the two-minute demo that starts at https://vimeo.com/38272912?share=copy#t=1731.06
2024-08-25 01:37:09 <xentrac     > well, a minute or two after that
2024-08-25 01:38:53 <xentrac     > it starts at https://vimeo.com/38272912?share=copy#t=1889.743
2024-08-25 01:39:02 <xentrac     > at 31'29"
2024-08-25 01:42:13 <xentrac     > yeah, so if you watch two minutes from there you will be ANGRY that our user interfaces don't work like that
2024-08-25 01:43:49 <xentrac     > there are actually prototype user interfaces that work like that going back to the 70s with GENESYS
2024-08-25 01:45:31 <xentrac     > this is Ron Baecker's GENESYS, a 16mm film that survives from 01970: https://www.youtube.com/watch?v=GYIPKLxoTcQ
2024-08-25 01:46:49 <xentrac     > it was pen-driven rather than multitouch or mouse-driven, because the only mice then were at SRI, and multitouch screens wouldn't be invented for another 30 years, but it uses the same thing Victor calls "performing with my hands", a sort of digital puppeteering
2024-08-25 01:47:46 <xentrac     > I'm not sure to what extent that kind of digital puppeteering is necessarily an improvement over sketching curves
2024-08-25 01:47:56 <xentrac     > neauoire: have you tried it?  you've done a bunch of animation stuff
2024-08-25 01:48:47 <xentrac     > the Object-Oriented Drawing demo I linked above is 5'33": https://www.youtube.com/watch?v=WgzdFoXLX9c
2024-08-25 01:49:39 <xentrac     > the first 30 seconds are a demo of what not to do though
2024-08-25 01:56:10 <xentrac     > zelgomer: all of these research prototypes have different attacks on the finger occlusion problem you're talking about, but the one that attacks it most directly is FingerGlass: https://www.youtube.com/watch?v=MNjTvz-nO_0
2024-08-25 01:57:20 <xentrac     > but for example AVNC (and apparently Termux-X11) have a "simulate touchpad" mode where you move your finger on the display to move around a mouse cursor that isn't near your finger, then tap to do clicks, which is already a huge improvement for me most of the time
2024-08-25 01:58:10 <xentrac     > and the Bret Victor demo I linked above uses a similar kind of indirect manipulation where he uses one hand to select the item he wants to manipulate from a list in a sidebar, while using the other hand to move it around
2024-08-25 01:58:27 <xentrac     > and I've seen games where you use your finger to move a piece around, but the piece floats above your finger so you can see it
2024-08-25 01:59:12 <xentrac     > the FingerGlass video is 3'41"
2024-08-25 01:59:51 <zelgomer    > c
2024-08-25 02:00:40 <neauoire    > xentrac: I haven't, just observed it from afar : )
2024-08-25 02:02:01 <neauoire    > I can't watch any video at the moment
2024-08-25 02:02:26 <neauoire    > we have very little bandwidth, there's a cell tower over a mountain nearby, but at low tide, we don't get much across
2024-08-25 02:02:44 <neauoire    > I'm looking forward to approaching a city so I can watch videos again
2024-08-25 02:03:51 <neauoire    > I don't like cameras filming stuff(folk.computer and dynamicland) sort of things
2024-08-25 02:04:00 <neauoire    > but I quite like graphical input methods
2024-08-25 02:04:09 <neauoire    > in fact, I use one nearly everyday
2024-08-25 02:04:10 <neauoire    > https://wiki.xxiivv.com/site/grail
2024-08-25 02:04:51 <neauoire    > it's not much but it allows me to have the wacom tablet over the thinkpad keyboard, and if I want to write notes or anything, I can use my little onscreen keyboard, which is more like a palm graffiti surface than a keyboard
2024-08-25 02:05:32 <zelgomer    > it looks neat but at the end of the day their solution is just make the ui bigger. it doesn't seem like something that would help much for typing, which is actually what i do a lot with my phone. i wish they would just make phones with keyboards again.
2024-08-25 02:06:35 <neauoire    > this url contains a copy of the rom if anyone is keen to mess with it
2024-08-25 02:06:36 <neauoire    > https://paste.sr.ht/blob/761b6f639b73b772978a5702ca6e1d119e3d0d54
2024-08-25 02:09:28 <xentrac     > neauoire: didn't you and Rek do a bunch of animation stuff on uxn?
2024-08-25 02:09:41 <neauoire    > yeah
2024-08-25 02:10:20 <neauoire    > but nothing like vector animation
2024-08-25 02:10:20 <neauoire    > it's more like traditional animation
2024-08-25 02:10:22 <xentrac     > I was wondering in particular if you'd done what I was calling "digital puppeteering", where the movement of an object in a scene is taken from the *recorded* movement of your hand
2024-08-25 02:10:52 <neauoire    > no, unfortunately none of our device have a working camera
2024-08-25 02:11:03 <xentrac     > recorded with your mouse or a tablet, not recorded with acamera
2024-08-25 02:11:05 <xentrac     > *a camera
2024-08-25 02:11:12 <neauoire    > Oh..
2024-08-25 02:11:23 <neauoire    > mhmm let me think
2024-08-25 02:11:37 <xentrac     > that is, where the *time* of the pointer motion is important
2024-08-25 02:12:04 <xentrac     > where your pointer is used to input an (x, y, t) curve rather than just an (x, y) point or curve
2024-08-25 02:12:41 <neauoire    > does this count haha
2024-08-25 02:12:41 <xentrac     > zelgomer: yes, I agree that a keyboard is a better keyboard than a touchscreen!
2024-08-25 02:12:41 <neauoire    > https://paste.sr.ht/blob/128a6e8483e47108069c261cf963a9f582618d3f
2024-08-25 02:13:00 <zelgomer    > lol
2024-08-25 02:13:06 <xentrac     > I don't know, maybe I can try it
2024-08-25 02:13:41 <neauoire    > orca whale puppeteering
2024-08-25 02:13:49 <xentrac     > I was going to mention (to unjust or whoever is looking at multitouch UI paradigms) that the last 45 seconds of Michael McGuffin's multitouch pie menu is also pretty inspirational: https://youtu.be/q97eTf8NbT0?si=5CmkDzAGqs35Bf5S&t=389
2024-08-25 02:13:58 <xentrac     > *pie menu demo is
2024-08-25 02:14:24 <neauoire    > I haven't had a phone for 9 years, I've missed out a lot of advance in touch device thingies
2024-08-25 02:14:43 <xentrac     > aha, I have the orca moving
2024-08-25 02:14:51 <xentrac     > but is there a way to record and play back its movement?
2024-08-25 02:14:59 <xentrac     > like to make an animation you could share with somebody else
2024-08-25 02:15:03 <neauoire    > there isn't unfortunately
2024-08-25 02:15:15 <neauoire    > it wouldn't be that much work to add tho
2024-08-25 02:15:49 <neauoire    > uxn roms are basically just images right
2024-08-25 02:16:04 <neauoire    > so when I make a rom, I just export the state of a program at a specific time
2024-08-25 02:16:19 <neauoire    > writing x,y keyframes in memory is pretty easy
2024-08-25 02:16:33 <xentrac     > right!  it's a question of designing the desired UX
2024-08-25 02:16:43 <xentrac     > and working on it until it works well for you
2024-08-25 02:16:59 <xentrac     > so I was curious whether you'd tried that paradigm and what you found had worked well or badly
2024-08-25 02:17:17 <neauoire    > https://paste.sr.ht/blob/12d14fcdecf70da1ff7f9eb733757fa45ba19ef2
2024-08-25 02:17:33 <neauoire    > this is something like the orca toy, but it records keyframes of the mouse positon
2024-08-25 02:17:33 <xentrac     > with respect to Grail, you might be already aware that one of the very first graphical user interfaces for programming was called GRAIL, using the only graphics tablet in the world at the time: https://www.rand.org/content/dam/rand/pubs/research_memoranda/2006/RM6001.pdf
2024-08-25 02:17:50 <neauoire    > oh yes of course
2024-08-25 02:18:05 <neauoire    > I wanted to use grail, so I just went and implemented it as closely as I could
2024-08-25 02:18:17 <xentrac     > yes!
2024-08-25 02:18:17 <neauoire    > my system use a few extra control bytes, so I added those
2024-08-25 02:18:37 <xentrac     > oh, you're using the algorithm RAND were using for handwriting recognition?
2024-08-25 02:18:46 <neauoire    > yup
2024-08-25 02:18:53 <neauoire    > https://rabbits.srht.site/decadv/
2024-08-25 02:18:56 <xentrac     > aha!  not a coincidence at all then :)
2024-08-25 02:19:03 <neauoire    > you can see here how I hook it up
2024-08-25 02:19:30 <xentrac     > oh cool, I still don't understand SECD
2024-08-25 02:19:56 <neauoire    > it's not really worth the trouble
2024-08-25 02:20:00 <neauoire    > it's a 9 opcode stack machine
2024-08-25 02:20:08 <neauoire    > well, spagetti stacks
2024-08-25 02:20:24 <neauoire    > it has a couple of lists in parallel for different parts of the environment
2024-08-25 02:20:28 <xentrac     > it might not be something I want to actually use, but it might be useful to understand it
2024-08-25 02:20:32 <neauoire    > opcodes are kept in the list as numbers
2024-08-25 02:20:46 <neauoire    > https://wiki.xxiivv.com/site/secd.html
2024-08-25 02:20:47 <neauoire    > try it out
2024-08-25 02:20:58 <neauoire    > I tried to make the c codebase as simple as I could
2024-08-25 02:21:08 <neauoire    > https://wiki.xxiivv.com/site/lisp
2024-08-25 02:21:23 <neauoire    > the lisp documented here is esp designed to be compiled for the SECD
2024-08-25 02:21:49 <xentrac     > that makes sense
2024-08-25 02:22:05 <neauoire    > the opcodes are kind of straight forward
2024-08-25 02:22:10 <neauoire    > the main issue is that it's SLOW
2024-08-25 02:22:16 <neauoire    > slow as all hell
2024-08-25 02:22:27 <neauoire    > even as far as lisps are concerned
2024-08-25 02:22:37 <xentrac     > that's a common problem with theoretical minimal machines :)
2024-08-25 02:22:39 <neauoire    > it's slower than Maude.. it's saying a lot
2024-08-25 02:22:45 <xentrac     > the prover?
2024-08-25 02:22:47 <neauoire    > yeah
2024-08-25 02:22:57 <xentrac     > I've never tried to do anything in Maude
2024-08-25 02:23:02 <neauoire    > it's fantastic
2024-08-25 02:23:13 <neauoire    > I use Modal nowadays
2024-08-25 02:23:26 <xentrac     > but maybe more fantastic as a prover than as an interpreter?
2024-08-25 02:23:26 <neauoire    > but Maude is well worth trying out
2024-08-25 02:23:38 <neauoire    > nha, it's just an all around awesome language
2024-08-25 02:23:48 <xentrac     > I was really depressed about the perormance of Bicicleta when I got it running
2024-08-25 02:23:49 <neauoire    > it cracks you brain in all sorts of fun ways
2024-08-25 02:23:56 <xentrac     > *f
2024-08-25 02:24:14 <neauoire    > Bicicleta?
2024-08-25 02:24:15 <neauoire    > what's that
2024-08-25 02:24:27 <xentrac     > because it's doing a naive tree walking interpreter thing with a naive alist environment
2024-08-25 02:24:41 <xentrac     > oh, well, I wanted my computer to be a bicycle for the mind
2024-08-25 02:24:54 <neauoire    > bicycles are hella complicated
2024-08-25 02:24:58 <neauoire    > stilts are more simple
2024-08-25 02:25:02 <neauoire    > you want stilts for the mind
2024-08-25 02:25:14 <xentrac     > bicycles go faster and are less tiring ;)
2024-08-25 02:25:25 <xentrac     > so I thought it would be good to figure out why programming was hard and fix it
2024-08-25 02:25:40 <neauoire    > there's always a tradeoff somewhere
2024-08-25 02:25:57 <xentrac     > and I decided that a big part of the problem was that you have to spend a lot of mental effort on figuring out what was going on
2024-08-25 02:26:05 <xentrac     > simulating the computer in your mind
2024-08-25 02:26:15 <neauoire    > did you write about bicicleta anywhere?
2024-08-25 02:26:23 <xentrac     > yeah, a lot on kragen-tol
2024-08-25 02:26:29 <xentrac     > also there's a Darcs repo at uh
2024-08-25 02:26:40 <xentrac     > http://canonical.org/~kragen/sw/bicicleta/
2024-08-25 02:26:45 <neauoire    > I'm surprised to hear that from you, you always seemed like a meet the computing more than halfway type of person
2024-08-25 02:27:31 <xentrac     > heh
2024-08-25 02:27:42 <neauoire    > oh shit ML
2024-08-25 02:27:51 <neauoire    > yeah well ...
2024-08-25 02:27:56 <xentrac     > sorry, that was the easiest way to get an interpreter running
2024-08-25 02:28:11 <xentrac     > ML is fantastic for writing compilers and interpreters!
2024-08-25 02:28:17 <neauoire    > it is!
2024-08-25 02:28:27 <neauoire    > but yeah, incrementing a number is like 9 billion cycles
2024-08-25 02:28:48 <xentrac     > OCaml can increment a number in two instructions
2024-08-25 02:28:52 <xentrac     > if you compile with ocamlc
2024-08-25 02:29:16 <neauoire    > ocaml is lot faster yeah
2024-08-25 02:29:17 <xentrac     > but incrementing a number in Bicicleta ended up being like 3000 instructions, slower than Tcl
2024-08-25 02:29:20 <xentrac     > slower than Bash
2024-08-25 02:29:31 <neauoire    > so what happened with bicicleta?
2024-08-25 02:29:34 <xentrac     > maybe not quite as slow as gdbscript
2024-08-25 02:29:36 <neauoire    > why is it so slow?
2024-08-25 02:29:57 <xentrac     > well, it's a naive tree-walking interpreter, but also all the operations are late-bound at a very deep level
2024-08-25 02:30:40 <neauoire    > ah yeah okay
2024-08-25 02:30:42 <xentrac     > and all its associative arrays (for closure-captured variables and methods) are linear linked lists
2024-08-25 02:31:07 <neauoire    > they're slow, but it's easy to get reasonable speed from the start
2024-08-25 02:31:20 <neauoire    > and you get high level like convenience with seemingly no schafolding
2024-08-25 02:31:28 <neauoire    > https://wiki.xxiivv.com/site/modal
2024-08-25 02:31:37 <neauoire    > modal has pretty much just one opcode
2024-08-25 02:31:37 <xentrac     > I don't know if you have a clone of dev3 but if not http://canonical.org/~kragen/sw/dev3/qfitzah.s and http://canonical.org/~kragen/sw/dev3/example.qf1
2024-08-25 02:32:04 <neauoire    > I don't, I think it'll fail if I try cloning this
2024-08-25 02:32:33 <xentrac     > oh, this looks very similar to http://canonical.org/~kragen/sw/dev3/sendat.py
2024-08-25 14:37:45 <user51      > I was googling around for whatever I could find about factoring, but I keep getting results about Factor.  Someone's trying to pull a Xerox  :)
2024-08-25 17:26:28 <xentrac     > heh
2024-08-25 17:39:19 <neauoire    > morning o/
2024-08-25 18:17:36 <xentrac     > morning!
2024-08-25 18:27:22 <neauoire    > for the longest time I shunned macros in uxntal but I've recently come around them and started to implement support for them in assemblers, linters, IDEs,..
2024-08-25 18:27:34 <neauoire    > kind of finishing up the work there today
2024-08-25 20:48:09 <veltas      > 'macros' expressions etc are essentially free in a Forth assembler
2024-08-25 20:48:27 <veltas      > Which is why I think they make good assemblers, in theory
2024-08-25 21:05:38 <xentrac     > yeah, I think even simple macros add a lot of expressivity and consequently DRYness to assemblers
2024-08-25 21:06:09 <xentrac     > and flexibility
2024-08-25 21:06:40 <xentrac     > but with flexibility unavoidably comes some bug-proneness and necessity for debugging
2024-08-25 21:07:34 <xentrac     > Forth subroutines are I think a better flexibility/unpredictability tradeoff than conventional text-subsitution macros like GPM or gasp
2024-08-25 21:07:40 <xentrac     > or the C preprocessor
2024-08-25 21:10:21 <xentrac     > but I suspect that term-rewriting like Modal or Maude (which you can also do in Coq, even though it isn't its primary approach: http://research.microsoft.com/en-us/um/people/nick/coqasm.pdf) has a better programming paradigm for that kind of thing
2024-08-25 21:10:30 <xentrac     > *is a better
2024-08-25 21:11:24 <xentrac     > partly on the theory that term-rewriting works really well for things like compiler optimization and code generation, and partly based on experience with C++ templates, which also embody a term-rewriting paradigm
2024-08-25 21:13:11 <xentrac     > you could see moving from procedural Forth to term-rewriting as sort of "going back to" text substitution, but I don't think that's correct in terms of the flexibility/predictability tradeoff
2024-08-25 21:13:47 <xentrac     > I feel like term rewriting benefits a *lot* from garbage collection and the Lisp object-graph memory model
2024-08-25 22:51:12 <neauoire    > I stayed away from macros because it made code that was harder to optimize
2024-08-25 22:51:21 <neauoire    > and I noticed it warped how people program and made bad code
2024-08-25 22:51:50 <neauoire    > but now that type checker is getting better and the linter smarter, it's starting to be more practical to do some specific things using macros
2024-08-25 23:12:55 <xentrac     > sometimes it makes code easier to optimize!  Like, in http://canonical.org/~kragen/sw/dev3/server.s, there's a .macro be near the end which is used to set registers to constants
2024-08-25 23:14:14 <neauoire    > ah yeah, but that'd be an anti-pattern in uxntal
2024-08-25 23:14:25 <neauoire    > the issue I had was that people were using macros to set constants
2024-08-25 23:14:28 <neauoire    > which is a really bad idea
2024-08-25 23:14:36 <xentrac     > so I can say `be $2, %ebx` to set %ebx to 2
2024-08-25 23:14:37 <neauoire    > (in uxn)
2024-08-25 23:14:58 <xentrac     > and that gets optimized into a three-instruction sequence which is shorter than `mov $3, %ebx`
2024-08-25 23:15:21 <xentrac     > is that the kind of thing that you mean?
2024-08-25 23:15:47 <neauoire    > sort of yeah, the linter now would throw a warning that these 3 instructions can be optimized
2024-08-25 23:16:04 <xentrac     > my objective here was to write a web server in under 2KiB of code, so that kind of thing was worthwhile
2024-08-25 23:18:13 <xentrac     > actually with the power of sstrip from ELFkickers (https://www.muppetlabs.com/~breadbox/software/elfkickers.html) the web server executable is 1830 bytes
2024-08-25 23:18:39 <neauoire    > that's nice, very small : )
2024-08-25 23:19:37 <xentrac     > with the standard binutils it had gone overweight when I recently added content-types to it for .css and .pdf
2024-08-25 23:20:10 <xentrac     > 2060 bytes
2024-08-25 23:22:05 <xentrac     > it's handy for things like copying files around the local network or testing things like http://canonical.org/~kragen/sw/dev3/raycast.html (a Wolf3D-style raycaster in 8033 bytes with a map you can edit with the mouse in real time)
2024-08-25 23:22:57 <neauoire    > I was wondering
2024-08-25 23:23:07 <neauoire    > do you still update dercuano.github.io?
2024-08-25 23:23:21 <xentrac     > no, dercuano is finished, immutable since publication
2024-08-25 23:23:37 <neauoire    > it's a treasure trove
2024-08-25 23:23:39 <xentrac     > see http://canonical.org/~kragen/dercuano
2024-08-25 23:23:43 <xentrac     > thanks :)
2024-08-25 23:24:04 <xentrac     > it has sequels: http://canonical.org/~kragen/derctuo and http://canonical.org/~kragen/dernocua
2024-08-25 23:24:10 <neauoire    > how large is the tarball?
2024-08-25 23:24:12 <neauoire    > I might grab it
2024-08-25 23:25:04 <xentrac     > the Dercuano tarball is 3.4MB
2024-08-25 23:25:18 <neauoire    > I'll grab it in the fall when we reach port
2024-08-25 23:25:19 <xentrac     > Derctuo is 8.6MB
2024-08-25 23:25:29 <xentrac     > hopefully it'll still be available
2024-08-25 23:25:32 <neauoire    > I've local copies of a few pages
2024-08-25 23:25:41 <neauoire    > but having the whole thing would be neat
2024-08-25 23:25:43 <xentrac     > Dernocua is 2.7MB
2024-08-25 23:26:42 <neauoire    > anything more than 500kb kicks me offine
2024-08-25 23:26:47 <xentrac     > yeah, the objective of Dercuano was to propagate as copies, but then sbp put up dercuano.github.io
2024-08-25 23:27:32 <xentrac     > the satphone provider only gives you 500 kilobytes a day?
2024-08-25 23:27:35 <neauoire    > I'll glady keep a copy
2024-08-25 23:27:38 <neauoire    > no no
2024-08-25 23:27:43 <neauoire    > it depends on the weather
2024-08-25 23:27:59 <neauoire    > but constant download seems to fail after a second or two of constant activity
2024-08-25 23:28:21 <neauoire    > and when it fails, I have a minute offline
2024-08-25 23:28:28 <neauoire    > it's kind of irritating
2024-08-25 23:28:32 <neauoire    > but we manage
2024-08-25 23:29:03 <xentrac     > oh, maybe try wget --limit-rate=20k http://canonical.org/~kragen/dercuano-20191230.tar.gz
2024-08-25 23:29:07 <xentrac     > that limits to 20 kilobytes per second
2024-08-25 23:29:17 <neauoire    > lemme try
2024-08-25 23:29:24 <neauoire    > that even fails sometimes
2024-08-25 23:29:41 <xentrac     > wget also supports -c to resume a partial download
2024-08-25 23:30:46 <neauoire    > yup, that's a trick I use a lot
2024-08-25 23:30:47 <neauoire    > mhmm
2024-08-25 23:31:19 <xentrac     > the web pages are small: http://canonical.org/~kragen/dercuano is 2132 bytes, http://canonical.org/~kragen/derctuo is 3102 bytes, and http://canonical.org/~kragen/dernocua is 1131 bytes
2024-08-25 23:31:29 <neauoire    > 60%
2024-08-25 23:31:42 <xentrac     > that's great!