Ethan Blanton / is on Nostr

I'm here for old computers, radios, and cool nerdery.

I identify with #Unix #Linux #HamRadio #retrocomputing and #making.

Professionally I am a computer scientist, now primarily an educator of college students. I am a #SystemsProgrammer with deep experience in #networking, #EmbeddedSystems, and #OperatingSystems. I have also worked in #security from time to time.

SDF Web: https://elb.sdf.org/
Alt: https://mastodon.sdf.org/@elb
Web: https://lostbits.net/elb/

Last Notes

2024-07-09 22:54:38

- reply

but but but ... it runs on macOS?!? While I'm honestly very glad that there are two thriving compiler toolchains, C++ in particular is troublesome between compilers.

2024-07-09 22:48:02

- reply

(Not directly related and not a suggestion, just musing about BSD support...)
I looked at this KiCAD/FreeBSD support issue the last time you mentioned it, and the patches against KiCAD 7 for ports are super trivial. I wonder what keeps this out of regular maintenance (or if it's just that KiCAD 8 is pretty new?).

2024-07-08 01:09:08

Some observations:
1. If one stupidly mounts the BeagleBone to the wrong side of a UniBone via ALL 92 SOLDER-THROUGH HOLES, it is POSSIBLE to rework it with a soldering station, hobby heat gun, and solder braid.
2. It is not REASONABLE to do so.
3. My back and neck hurt terribly.
4. I am out of solder braid.
Draw your own conclusions. #retrocomputing #electronics #unibone #pdp11

2024-07-05 22:18:37

- reply

That's fine, and reasonable, although existing programs may expect to be able to move the cursor down without returning it to the left. This is why the Unix console character driver emits \r\n when \n is printed to the console in cooked mode.
Of course, when it was written, the paper had to be fed and the carriage had to be returned. (And that's why it's CR _then_ LF, because CR was often slow, and that gave it an extra ten bits to get there!)

2024-07-05 21:09:07

- reply

what do you do if you reach the end of the line and there's no newline? I think that answers this question, too - and there's more than one reasonable answer, but I think auto-generating a cr/lf before column 81 is what I would go for. That puts the next tab stop at the first stop of the next line.

2024-07-05 19:45:41

- reply

It feels fair for the ROM terminal emulator just Be 80x24, and if someone is writing their own terminal emulator they can encode their own tab logic. ;-)
I would personally use one bit per column for tab stops, unless you anticipate needing other column attributes to justify the full byte. Many terminals supported a bell upon crossing a column, but that's because they printed on paper and would just hammer a hole in the paper at the right margin if you didn't CR.

2024-06-25 15:54:21

- reply

In general I am NOT a fan of reversed footprints like that, it's awfully easy to put them both in the same way. Your silkscreen symbols are fairly obvious, but still, tough to tell from the backside when you're laying down the solder.

2024-06-14 02:50:24

- reply

When I'm doing rapid kernel development I normally bring up a serial loader of some kind, and then just put some trampolines to well-known RAM locations in ROM (I think you're planning to use trampolines CP/M / MS-DOS style anyway, right?). At boot it triggers the serial loader, you blast your working code into RAM, and go from there. If _most_ of your code is reasonably position-independent, this works well.

2024-05-28 02:53:27

- reply

I feel like the footprint for the little reset regulator guy might be a little undersized, if everything is to scale there.

2024-05-22 04:52:06

- reply

vias aren't great for current carrying, either, and they're inside the board, which means there's nowhere for the heat to go.
I expect the right answer to be fat traces on the component side of the board that neck down where they have to to fit, for the shortest length possible. The nearby fat copper traces will conduct the heat away effectively, particularly if you use a pour or just oversize them.
That's the same way the component pads get away with being so small.

2024-05-09 19:23:32

- reply

I use a lot of ZFS and a lot of LVM, and I ... don't really agree that LVM gives you "most of" what ZFS does, except in marketing bullet points. I like LVM, but the ZFS proposition is pretty different in a lot of ways. For one, provisioning works _radically_ differently, and for another, recovering ZFS after an unexpected drive failure is much less fiddly. That's not to mention the verifiability aspect.
I agree that LVM is great, and I use it on my desktop, but my NAS/backup is ZFS.

2024-04-20 02:01:07

- reply

Rob Pike agrees: https://harmful.cat-v.org/cat-v/

2024-04-15 01:25:31

- reply

1 is more of a meme among non-Emacs users than true, 2 I simply don't agree with (although I do understand that there are absolutely complexities with GNU licensing, as both an open source author and user), 3 ... is pretty hard to swallow, however problematic Stallman is. In particular, gcc and binutils absolutely paved the way to the world of usable computing we have today, even if you don't care for Emacs. I wouldn't invite the guy to my events, but I cannot throw out Emacs or gcc.

2024-04-14 22:42:54

- reply

I know I sound like a broken record when you bring these things up, but ... that is exactly what Emacs is! And it can do it for a half a dozen or more languages (some of which cannot easily output compiled native applications, but as I recall, neither could the free QBasic that came with DOS). The editor-is-my-repl-and-IDE paradigm fits very cleanly with most LISPs and several modern interpreted languages like Python and Lua, and is also popular with ML/Haskell/etc.