Weekly Head Voices #218: Back to the future


Welcome to the Weekly Head Voices #218, looking back at the week from Monday March 22 to Sunday March 28 of the year 2021.

Figure 1: A little forest on Vergelegen farm, close to where we had just enjoyed an absolutely fabulous Sunday lunch, with wine.

Figure 1: A little forest on Vergelegen farm, close to where we had just enjoyed an absolutely fabulous Sunday lunch, with wine.

Two worlds collide!

On the one hand I do worry about the number of hours I spent on this, but on the other hand it’s one of those time investments that should pay off handsomely over the coming years.

My dream is that one day we’ll have an organizational knowledge management system in place that will enable the almost effortless creation, maintenance and especially the effective retrieval of relevant documentation.

Because the organization in question lives in the Microsoft ecosystem, the first laborious steps I have been taking were figuring out the various ways of getting technical documentation from the hands of engineers all the way into the company SharePoint.

(I hope you can forgive me, Charl of the past… I do have my reasons.)

Here I would like to share just one of the many dirty tricks I’ve learned so far: The easiest way to host a static (Hugo) web site on SharePoint, and hence having it fully indexed by Microsoft Search, is simply to dump the whole thing into any modern site’s Site Assets library, but only after renaming all html files to aspx.

As soon as I can get PnP PowerShell to push updates to full sites a bit more quickly, I’ll hopefully be able to hook it up to Azure pipelines.

Engineers write markdown, documentation magically appear on company SharePoint!

Let’s keep it compact

On Thursday November 14, 2019 at 17:48 I made a mental note, where by “mental” I mean “in my Emacs”, to prefer Zstandard (or zstd) compression for everything, based on a number of posts I had read when I was actually intending to look into when xz should be preferred over bz2.

In the meantime, zstd’s compression ratios, along with its multi-threaded compression speed and especially its decompression speed, have helped me more than once, which is why I’m mentioning it here.

Here’s the 2 second guide:

# compress using as many threads as there are CPU cores
tar cf - giant_directory | zstd -T0 -o giant_directory.tar.zst
zstd -dc giant_directory.tar.zst | tar x

You can even get zstd to train on your type of data in order to build up a domain-specific compression dictionary!

Can type systems catch bugs? Science says yes!

Thanks to this forum post by Araq, the creator of the Nim programming language, I read this conference paper describing this really neat research project where the researchers retroactively type-annotated a random selection of publically available bugs in un-typed JavaScript code, and then showed that about 15% of of those bugs would have been caught.

In their words:

Evaluating static type systems against public bugs, which have survived testing and review, is conservative: it understates their effectiveness at detecting bugs during private development, not to mention their other benefits such as facilitating code search/completion and serving as documentation. Despite this uneven playing field, our central finding is that both static type systems find an important percentage of public bugs: both Flow 0.30 and TypeScript 2.0 successfully detect 15%!

I found the simple but quite solid experimental methodology in the paper delightful, over and above the fact that this is a great one to have in my database for arguments about the utility of type systems.

(This would possibly have made dynamic-type-slinging Charl from the past go red in the face.)

Science doesn’t give a s***t about my beliefs either

That makes for two (2) points in this short blog post that would have caused Charl from the past to get quite unhappy.

That’s science for you folks: Reality develops, or it just stands there, really really still, and/or observations yield new data, and then all of your models have to be updated!

Forerunner hot takes

The more-or-less free Apple Watch 3 that I acquired in July of 2018 has served me really well over the years, but for mysterious reasons that I can’t explain, the time had come to upgrade to a real running watch.

After weeks of reading reviews, creating little comparison tables and complaining to friend LM about the utterly devious way in which Garmin has segmented their market into tiny little pieces with frustrating borders, I finally settled on the Garmin Forerunner 245 (without music).

I made sure that the watch would be delivered this past Friday so that I could take it for a spin or two over the weekend.

My first observations are:

  • One really notices all of the Apple polish when you have to give it up. It’s clear that the Forerunner is a robust tool for running, but heck is it klunky. Every time I push one of the four physical buttons, I imagine the watch going “hooooonk!”.
  • The device includes a whole bunch of advanced-looking new-to-me metrics by FirstBeat (recently acquired by Garmin) like Stress measurement, the Body Battery (taking into account sleep and stress to predict how much energy you should have), advanced recovery time (this is pretty useful: how many hours before my next run, based on how hard the previous one was, stress levels and also on sleep quality in the meantime) and so on.
  • I now have an optical pulse oximeter on my arm, which, although far from accurate, does give at least some estimate of my blood oxygen. This is important to me especially to ensure that that aspect of my sleep is also going as it should.
  • About that sleep tracking… the sleep onset detection must be a bit of an embarrassment for at least some of the folks at Garmin. For example, last night the watch detected onset of sleep while it was lying on the bedside table during my pre-bedtime shower. Surely with that HR and PulseOx sensor, the watch should know that it’s not even on my arm? Apple Watch 3 does this with 100% accuracy with only its HR sensor. (To add insult to injury, the Garmin detected that onset an hour before my configured bedtime.)
  • On the other hand, the battery life on this thing is, as is to be expected from a Garmin, quite phenomenal. (this is probably partly due to all of the half-implemented algorithms haha)
  • More generally speaking, I’ve disabled most of the notification features, so that the thing on my arm is more like an old-fashioned, mild-mannered watch that transforms into super-watch when running-related activities have to be undertaken. That in itself is an improvement that is not to be sneezed at.
  • In summary: Acknowledging the multitude of biases at play, I am really enjoying the new device. In addition, my Apple Watch 3 could hereby enter the family pool, freeing up another AW3 which now lives on the proud arm of my first running GOU.


Thank you folks for reading these words, and especially for reconstituting thoughts that are probably similar to the ones that happened at my end before and during the production of said words.

You’re more than welcome to add some words of your own in the comments section below. (I literally installed and configured it for you. No pressure though. You know me.)

On that note (specifically the note of adding your own words): A part of my brain has been entertaining the idea of setting up a discord server called The Church of Love and Science and/or Friends of Charl.

The brain part in question claims that it would be a place for friends, and for any other fans of Love and Science, to hang out and sometimes make interesting new connections, and to enjoy the existing ones.

What do you think?

Err, let me know in the comments, or via Signal, or via discord?