Tried doing some electronics this morning. Not much to show for it apart from some reading, designing, and driving to Jaycar to grab some components. This really is an activity where I’m still quite a novice.

I’ve been thinking of OS/2 recently. Yes, the ill fated OS that IBM built with Microsoft. Re-reading the Ars Technica write-up of it and listening to the Flashback episode again fills me with nostalgia.

Truth is that a lot of my early experiences with computing began with OS/2. Dad was working somewhere that used it, and I had a chance to play with it whenever he bought his laptop home. We had a plain old DOS home computer as well but it wasn’t as powerful or exciting as the laptop as dad’s laptop. It also helped that the laptop had a colour screen.

There were a lot of firsts associated with this OS. It was the first time using a graphical-based OS. I could do some pretty basic things in the DOS command line at the time, but being able to manipulate things on screen with a mouse was a much better experience. The OS had some pretty nice utilities that I grew to love, like a music app that let you compose monotonic songs by moving sliders that adjusted the pitch and duration of a note. It was my programming environment as well (QBasic, but it was still running on OS/2).

It was my first experience “shutting down” a computer. This was quite novel, and I remember wondering why it was necessary. The DOS computer you could just turn off, why couldn’t you do that here? I came to accept how important shutting down the computer first was, for no reason other than that not doing this would mean the next boot-up would run a file system check that would take several minutes. For someone who wanted to get to my DOS games as quickly as possible, learning this was important.

It was even the first time I saw someone surfing the web. On night I was watching dad using WebExplorer. I asked him what he was doing and he said that he was “using the web”. I had no idea what that meant, but I wondered if it had anything to do with a family friend that had the surname Webb. I still remember the loading animation of that browser: cubes flying by on the screen of the computer icon in the toolbar while the disk activity light flashed green.

When Windows 95 came around, we set it up with a dual boot system with OS/2. I remember at the time being reasonably unimpressed with Windows 95 and was more than happy to continue using OS/2, at least while we still had v2 installed. But eventually we got version 3 (OS/2 Warp) and it was around this time my love for it was starting to wane. DOS games were always a little incompatible on OS/2¹ but they eventually stopped running altogether, and I found myself booting into Windows a lot more often to play them. Eventually the day came when Dad bought home a laptop that only had Windows 95, and my OS/2 experience came to an end.

Anyway, it would be nice to pay around with OS/2 v2 again. Apparently OS/2 v2 is notoriously difficult to virtualise so I don’t know if that’s even possible. I found an emulator of OS/2 v1 that runs in the browser, and the GUIdebook has some fantastic screenshots of v2. I guess that will have to do for the moment to bring back the memories.

Follow-up to my last post: you can turn off committing after conflict resolution in GoLand by clicking “Modify options” in the Merge dialog and selecting “Do not commit the merge result.” This should display the --no-commit flag.

Ok, I admit it’s pretty close to how it works on the command line; although not when you have conflicts, and the only time I merge in GoLand instead of the command line is when I need to resolve conflicts (I prefer the diff/merge tools there). But to be fair to GoLand, it doesn’t know about my merging preferences.

I don’t understand code editors that think they know when to commit changes “better” than I do.

GoLand is guilty of this. I just finished resolving merge conflicts, but instead of letting me make sure the resolved conflicts actually build and pass the test, GoLand just commits them as soon as I’ve finished picking the hunks I want to use. This means any fixes I need to make cannot go into the merge commit. That commit is now unstable.

One could argue that this is the proper way to do things, that the merge commit should only contain conflict resolutions and nothing else. I don’t agree. I’m bound to make mistakes while resolving conflict, and I want to make sure what I’m committing is actually working code. It probably doesn’t matter in the grand scheme of things, and it wouldn’t be the first time I make a commit with dodgy code. But I’d like to reduce the number of times that happens if I possibly can.

Besides, conflicts in Go imports or module versions I can usually fixed by running tools. I tend to resolve these conflicts quickly, expecting to get errors and duplicates that I can fix with a couple of commands (they usually need to be formatted anyway). I can’t do this if the code editor decides when to actually commit the changes.

Stage the resolved conflicts all you like, but let me actually commit them when I say so.

It kills me that Pinboard doesn’t automatically set the title of a bookmark if you add one without it. So easy to fetch the page and get it from there. It can be a setting if privacy/costs are a concern. But really, having this feature would be a massive usability improvement.

Premature abstraction is the root of unnecessary rework.

Might just have to accept that the way I manage links to posts of interest is not perfect. There might be some room for improvement — something that polls Pinboard occasionally and posts any new links from there somewhere — but a wholesale replacement is probably not a good idea.

A reminder to myself that the only way to get a blog post out there that you’re happy with is get a blog post out there that you’re not happy with. Classic case of learning by shipping. And I think you need to actually need see it all the way through to publishing it. I’m not convinced that you’ll get the full benefit if you just leave the post half-finished in your drafts folder.

It’s been a rare (but not unprecedented) three coffee and one caffeinated tea kind of morning today. 😴

The cafe I go to has started playing music where I sit. I find music at a cafe distracting, mainly because I find myself paying attention to the music instead of what I was doing. Fortunately it’s music I don’t find appealing. But even so, bad music is still not no music.

I never considered myself someone who believed that Go must’ve had generics from the start. I appreciated that the designers added them to the language, but I though I’d could be just as effective writing Go code if they chose not to.

I don’t believe that anymore.

I’ve found generics in Go to be a major improvement to the language. It now means that I can now use higher order functions that operate on collections, like “map” or “filter”, in a more natural way. I got use to these functions while working in languages that had them (Python, Ruby, JavaScript, Java 1.8) and they’ve been so useful that I wished Go had them as well.

There was no technical reason for Go not to have these functions, but they would have had to use the interface{} type to be useful, meaning that there was no type safety built-in and your code would be littered with type assertions. This was such a turn-off that most of the time I didn’t bother with considering higher order functions, and just wrote a for loop by hand to convert types in one slice to types in another slice. Trivial to write, but just so boring.

Now with generics, these higher order functions can be made type safe, and there is no longer a need for type assertions to use them. These made them viable once more, and I’ve found myself using them a fair bit recently. Not having to write yet another for lop has made coding fun again.

Saw the following quote while reading this article:

You could fill a book with all I know, but with all I don’t know, you could fill a library.

— Unknown

Quite profound.

It’s 2023 now and podcasts are still saying that I should Google so-and-so to see an article they’re citing, instead of just putting a link in the show-notes. I understand that this is mainly because Apple’s podcasting app doesn’t support HTML with links. Maybe once they’ve stopped burning money on their podcast subscription play that seems to be going nowhere, they can take a look at fixing this. We’ve got linking technology now; we shouldn’t be afraid to use it.

On a train now peeking at someone’s laptop. His desktop is completely covered in document icons. And I mean completely, as in there’s no room for any more.

I’ve never seen this before. It’s both amazing and shocking at the same time.

Busses should use the tram lines more. I remember being taught that if a bus was driving on a tram line, it must be treated like a tram. This gives them dedicated lanes and certain priorities over cars. Don’t know why busses don’t do this.