[HN Gopher] KnightOS was an interesting operating system
___________________________________________________________________
KnightOS was an interesting operating system
Author : akalin
Score : 227 points
Date : 2020-01-27 23:47 UTC (23 hours ago)
(HTM) web link (drewdevault.com)
(TXT) w3m dump (drewdevault.com)
| geek_at wrote:
| At first I confused it with Temple OS. That was also very
| interesting but maybe less useful
| SirLotsaLocks wrote:
| honestly impressive, but yeah not very useful.
| coolreader18 wrote:
| Relevant: Ndless[1], a jailbreak for TI-Nspire calculators
| (Nspire {CX ,}{CAS,}). I've played around with the source code[2]
| as well as compiling Rust[3], Nim[4], and Zig[5] to target it.
| It's all pretty neat, there's a gba emulator for the TI-Nspire
| floating around somewhere as well.
|
| [1]: https://ndless.me
|
| [2]: https://github.com/ndless-nspire/Ndless
|
| [3]: https://github.com/coolreader18/rsspire
|
| [4]: https://github.com/coolreader18/Nim/tree/tinspire
|
| [5]: https://github.com/coolreader18/zigspire
| unlimitedbacon wrote:
| When I was in high school, I always wanted to build a better OS
| for the TI-84, but I did not have the skills or attention span to
| do it on my own. When KnightOS came along, I got involved and was
| able to take on individual features or issues as bite size
| chunks. I was able to make real contributions to a project that
| resonated with me on a fundamental level. In the process, I
| learned z80 assembly, how OSs work, but also most importantly how
| to contribute to a real open source project.
|
| Thanks, Drew.
| ddevault wrote:
| Hey, I remember you :) thanks for all of your help back in the
| day. I was just thinking to myself earlier that your
| improvements to the file manager added a lot of much needed
| polish.
| cjbprime wrote:
| Neat. Would be interested to see a quick video/GIF tour.
| ddevault wrote:
| Here you are:
|
| https://yukari.sr.ht/knightos.webm
| mintplant wrote:
| What's yukari.sr.ht? Personal file dump?
| ddevault wrote:
| Aye. It has a sr.ht subdomain for historical reasons, but
| it's a personal server.
| dmix wrote:
| > In fact, the "userspace" task switching GUI would read the
| kernel's process table directly to make a list of running
| programs.
|
| This is great. A perfect learning project is trying to build
| something yourself, then learning how other developers solved the
| problems or larger OSes handle it (or working backwards from that
| point and simplifying).
|
| I noticed some "#ifdef FLASH4MB" in the linked source code, was
| it difficult to update it for 4mb or were similar constraints in
| place where it was basically the same with more memory?
| cesarb wrote:
| Interestingly, "read the kernel's process table directly to
| make a list of running programs" is exactly how the "ps"
| command worked in the distant past. It's a bit hard to find
| information about this nowadays, but I could find a couple of
| SE answers (https://unix.stackexchange.com/a/340893 and
| https://unix.stackexchange.com/a/511530) which mention this
| fact.
| ddevault wrote:
| Nah, it was reasonably easy. The main issue with 4MB calcs is
| that it used a secondary hardware port to add two extra bits to
| the page address. Until then we were able to page with only one
| port, so everywhere was hardcoded to use it. We had to replace
| all of that with a marcro, and update a few places manually.
| There's a comment in that code which speaks to this:
| ; we can't use getBankA because it behaves differently on the
| 84+CSE, and we don't want that
|
| getBankA was the name of the macro we used to interrogate which
| Flash page was swapped into bank A (address space
| 0x4000-0x7FFF).
|
| The real pain with that port was the new color screen.
| papito wrote:
| Obstacles and difficulty make us better at things. Lack of
| obstacles makes us complacent. This is a known phenomenon.
|
| https://www.ted.com/talks/tim_harford_how_frustration_can_ma...
| kragen wrote:
| KnightOS is a system I've been wanting to try for years, although
| I don't have the hardware for it. It's an admirable achievement.
|
| How hard would it be to get KnightOS to be able to self-host?
| Something like an 8080 or Z80 is probably the wimpiest processor
| that it makes sense to try to build a self-hosted development
| environment on; although there were compilers that could run on
| the PDP-8 (address space: 4096 12-bit words), John Cowan tells me
| that in practice people did most software development for the -8
| on bigger machines like a PDP-10.
|
| CP/Mish is the other free-software operating system for the Z80
| (and also runs on the 8080), but it's pretty far from self-
| hosting. There's a free-software C compiler for CP/M which can
| run under CP/Mish (BDS C) but its dialect of C is not the one
| CP/Mish is written in.
|
| A really small bootstrapping environment like this could go a
| long way to making computers comprehensible and making the
| Karger-Thompson attack infeasible.
| ddevault wrote:
| Last time I checked, the source code can't even fit on the
| calculator's Flash chip.
| Koshkin wrote:
| There goes my day.
| zozbot234 wrote:
| With no MMU or hardware supervisor mode, this would be more of a
| simple task executor than an OS in any reasonably-modern sense.
| Still impressive though.
| dTal wrote:
| Also interesting: PunixOS, a sadly-dead project to port 4.4BSD to
| Motorola 68k calculators like the TI-89. Not just a Unix-alike,
| but a Real Unix (tm), with pre-emptive multitasking and
| everything.
|
| http://punix-os.blogspot.com/
| fsiefken wrote:
| interesting, seems similar to fuzix http://www.fuzix.org/
| cr0sh wrote:
| I was about to post about the TI-92+ not getting any love, saw
| this, got happy, then saw it dead...sad again. Oh well. :)
| msla wrote:
| > Not just a Unix-alike, but a Real Unix (tm),
|
| Not unless they paid The Open Group to use the very real Unix
| trademark.
|
| There's a part of zOS that's Real Unix (tm) and FreeBSD is not,
| so can we give it a rest?
| boring_twenties wrote:
| No one gives a shit about the Open Group's trademark.
| Seriously.
|
| 4.4BSD (unlike FreeBSD, or 4.4BSD-Lite(2)) contains
| proprietary code from AT&T Unix, therefore it's Unix in most
| people's eyes.
| msla wrote:
| > 4.4BSD (unlike FreeBSD, or 4.4BSD-Lite(2)) contains
| proprietary code from AT&T Unix
|
| Only three files had to be removed.
|
| https://en.wikipedia.org/wiki/UNIX_System_Laboratories,_Inc
| .....
| kick wrote:
| FreeBSD and 4.4 BSD are not the same thing. You seem bitter.
| Why are you bitter about something like that?
| msla wrote:
| FreeBSD is the modern 4.4BSD descendant.
| dang wrote:
| Can you email hn@ycombinator.com so we can send you an invite
| to post that? Invited posts go into in the second-chance pool
| (described at https://news.ycombinator.com/item?id=11662380).
| jacobush wrote:
| Also, Lunix (C64)
| sixothree wrote:
| I feel like constraints are where developers get to shine. On a
| system with (relative) unlimited resources, many developers seem
| to be able to complete the same tasks. But when there is scale
| involved or some constraint (like 4 mb of storage) you get to
| find out who is willing to learn what they need to learn to get
| the job done.
| reaperducer wrote:
| _I feel like constraints are where developers get to shine._
|
| Absolutely. And I think that's one of the ways you can tell
| that there is an art to programming. It's not just pushing
| buttons in the right order. (Something my boss doesn't
| understand because she took a computer class in high school and
| thinks she can do my job. "I know code.")
|
| Computer science is a thing. Programming is a thing. They're
| not the same thing. And as we get into more abstractions, I
| feel like the programming side is becoming more an art form
| than the science it is presumed to be.
| dec0dedab0de wrote:
| I always liked how rms says that programming is a craft.
| 0x445442 wrote:
| > Computer science is a thing. Programming is a thing.
|
| Engineering is also a thing but that's vanishing too.
| 1-6 wrote:
| Google Sheets is my constraint. Utilities.sleep()
| thedanbob wrote:
| My day job is web development but I love hacking on
| microcontrollers in my spare time. One project involved serving
| a static webpage from an ESP-8266. But how do you server an
| HTML file without a filesystem? There's probably a better way,
| but my solution was to gzip the file and embed it as an array
| of raw bytes within the program.
| Koshkin wrote:
| As far as program resources go, on DOS, at least, it used to
| be possible to simply append a file to the executable, and
| the loader would put everything in RAM. (On the other hand,
| you do not need a generic web server to serve a single static
| page; then you don't need a file system to store one file.)
| ronsor wrote:
| You can do that with most executable formats: MZ/NE/PE
| (DOS, Windows), ELF (Linux, BSD). Not sure about mach-o
| (macOS) though.
| emily-c wrote:
| This is still pretty common for things like self extracting
| archives on Windows. copy /b
| "%windir%\system32\extrac32.exe"+inputarchive.cab
| selfextract.exe
| wolf550e wrote:
| Do you mean like Apache mod_asis [0] which will serve a pre-
| formatted HTTP response, including the headers, directly from
| disk?
|
| Then your HTTP server can speak very little HTTP (compare
| request line to preformatted "GET /file HTTP/", return 404 on
| mismatch, return preformtted response on match). Maybe also
| return nice error if request does not have accept-encoding:
| gzip.
|
| 0 - https://httpd.apache.org/docs/2.4/mod/mod_asis.html
| GTP wrote:
| ESP8266 has basic filesystem support, have a look at this:
| https://arduino-
| esp8266.readthedocs.io/en/latest/filesystem..... You define
| the filesystem's size when you flash the ESP and there's a
| tool to upload files there. You should also be able to find
| example projects that serve a webpage read from SPIFFS. Edit:
| this is a fake AP for esp8266 that I wrote some time ago that
| has a captive portal that loads pages from SPIFFS if they're
| there, otherwise sends some default pages directly from code
| https://github.com/GTP95/CaptivePortalMatematico?files=1
| thedanbob wrote:
| Cool! I ended up switching to a raspberry pi for that
| particular project but I'll keep that in mind.
| deckard1 wrote:
| https://github.com/me-no-dev/ESPAsyncWebServer#respond-
| with-...
|
| That whole ESPAsyncWebServer github project has a lot of
| great examples on ways to use the esp8266.
|
| Also, if you want to connect to your esp8266 externally, you
| can setup nginx (maybe on a raspberry pi?) as a reverse proxy
| with TLS enabled: https://jjssoftware.github.io/secure-your-
| esp8266/. Probably a bit pointless if you only have one
| esp8266 to connect to, since the RPi could probably do it
| all. But if you have multiple IoT devices to connect to,
| might be worth it.
| invalidOrTaken wrote:
| The nice thing about constraints is that they know what they
| want.
| quotemstr wrote:
| That's a very pithy way of putting it
| jlg23 wrote:
| It's the litmus test: How much of taught abstractions do you
| really understand (can you adhere to the underlying coding
| practice without spelling it out in (costly) code? Do you
| really understand a runtime environment and can you make your
| basic tool, the compiler, optimize for your use-case?
|
| Back then, when I was young and stupid, I agreed to develop a
| Command & Conquer like game in j2me with a deliverable size of
| max 120k. I did, but that required lot of all-nighters and a
| severely bruised and corrected ego..
|
| And, as the senior dev, I got to listen to very smart recently
| graduated devs who did not know shit (as I did when I started).
| Best management lesson ever...
| Someone wrote:
| As Goethe wrote: _"In der Beschrankung zeigt sich erst der
| Meister"_ ("the master shows himself through self-limitation").
|
| One can see this as a form of constrained writing
| (https://en.wikipedia.org/wiki/Constrained_writing)
|
| Demos (https://en.wikipedia.org/wiki/Demoscene) and The
| submissions to the International Obfuscated C Code Contest
| (http://ioccc.org/) also are examples of this.
| SkyMarshal wrote:
| There are a bunch of styles of X-Driven-Development, and I
| honestly think we should add the notion of Constraint-Driven-
| Development as a first-class citizen to that mix. CDD.
|
| Start by specifying the constraints, including hardware,
| software, formal specification, and business rules. Put them
| all into a Constraints Document. Then tell the engineering and
| design teams to work together to make the coolest, most
| creative product they can, without violating the specified
| constraints. Use Agile, rapid-iteration methodologies if you
| like, within that framework.
|
| Then evaluate the outcome, and decide whether to iterate by
| updating the constraints document and repeating the process. Or
| maybe you're happy with v1.0 and no iteration necessary.
|
| Constraints drive both innovation and quality. I wonder if the
| process can be viably formalized in such a manner.
___________________________________________________________________
(page generated 2020-01-28 23:00 UTC)