Mastodon instance comparisons

May 26, 2025 at 10:43 AM by Dr. Drang

I forgot to mention in yesterday’s post another way (apart from looking at the FediDB directory) to check how much an instance is being used. That’s the activity API call. Running it on Fosstodon this morning via

curl -s https://fosstodon.org/api/v1/instance/activity | jq

I got

[
    {
        "week": "1748188656",
        "statuses": "1383",
        "logins": "2663",
        "registrations": "1"
    },
    {
        "week": "1747583856",
        "statuses": "16416",
        "logins": "6996",
        "registrations": "5"
    },
    .
    .
    .
    {
        "week": "1741535856",
        "statuses": "21653",
        "logins": "7844",
        "registrations": "24"
    }
]

The result was an array of 12 JSON objects, each with a week’s worth of activity.

1 The week entry for each item is the Unix timestamp for the beginning of the week. Converting the first week value to a human-readable form through

date -jr 1748188656

I got

Sun May 25 10:57:36 CDT 2025

so the first set of activities represents only about a day, not a full week.

Running

curl -s https://fosstodon.org/api/v1/instance/activity |\
jq '.[1:].[].logins | tonumber'

which omits the number of logins for the partial week, returns

6996
7140
7212
7622
7498
7440
7490
7516
7666
7882
7844

So the activity on Fosstodon has gone down, and some of that may be due to the recent controversy. Let’s check this against Hachyderm, an instance that I’ve always considered comparable to Fosstodon. Here,

curl -s https://hachyderm.io/api/v1/instance/activity |\
jq '.[1:].[].logins | tonumber'

returns

6531
6539
6651
6653
6579
6631
6735
6731
6786
7187
7223

so its logins have also gone down a bit over the past several weeks. Plotting both timelines, we get

Login timeline comparison

So both instances have had an overall decline in logins over this period. The short-lived jump in Fosstodon logins four weeks ago might well have been due to the controversy.

Another way to compare the two sets of data is to plot the ratio of logins:

Ratio timeline

There’s about a 7% variation in the ratio over this period, but it’s hard to argue for a trend. Certainly not an exodus from Fosstodon.

I should mention that “Weeks ago” is an imperfect scale because Fosstodon and Hachyderm start their weeks about a day apart. As we saw above, Fosstodon starts its week on Sunday morning (Chicago time); Hachyderm starts its week on Monday morning.

I suppose I could have compared the statuses data instead of logins, but I care more about the number of people checking their timelines than how much they’re posting. What’s nice is that the API gives you the information to compare either one.

  1. The docs say it returns three months worth of activity, so presumably sometimes it will return 13 objects instead of 12. ↩

Staying with Fosstodon

May 25, 2025 at 9:04 AM by Dr. Drang

A few weeks ago, I expressed some concern about staying on my current Mastodon instance, which is Fosstodon:

I guess it’s time to change instances once more. I’m going to spend a little time trying to decide on a new place because I really don’t want to have to do this again in another few years.

The Fosstodon moderator whose comments (on other social media) had gotten people upset left his position, as did another administrator. Still, several people left for other instances, and there was talk of Fosstodon being defederated. All of this happened before I was aware of it—FOMO? what’s that?—and the new admin team was already being built. But the defederation talk concerned me. Even if the wrongs were righted, that doesn’t help users if their posts aren’t being read elsewhere.

The simple thing would be to switch to mastodon.social, the biggest Mastodon instance. If it goes down, that’s pretty much the end of Mastodon, so there wouldn’t be any need to switch again. Unfortunately, the “drdrang” username wasn’t available there. Apparently, someone had signed up for it and had their account suspended, so the name wasn’t available to me.

With this path closed, I decided to be a little less reactive and look into whether Fosstodon was actually being defederated or not. I visited the server directory at FediDB and explored its list of Mastodon servers ordered by monthly active users. Fosstodon was still in the top ten, which suggested that there hadn’t been a mass exodus. But this didn’t check for defederation directly.

Each Mastodon instance has an About page,

1 and most of those pages include a list of “Moderated servers,” instances that have been suspended or limited in some way by the host server’s administrators. Fosstodon didn’t appear in the Moderated list of any of the main instances I checked, although some didn’t publish a Moderated list.

Then I tried out the peers API call, which returns a JSON array of all the servers the host communicates with:

curl -s https://mastodon.social/api/v1/instance/peers |\
jq '.[]' | fgrep fosstodon.org

This returned

"hub.fosstodon.org"
"fosstodon.org"

A little more clever use of jq was

curl -s https://mastodon.social/api/v1/instance/peers |\
jq 'any(.[]; . == "fosstodon.org")'

which yielded just

true

because one of the peers was just “fosstodon.org” with no other text. I tried this same call for several instances, and they all returned true.

Just to be sure that I understood what peers was returning, I tried it out with a few instances I saw on mastodon.social’s Moderated list. For example,

curl -s https://mastodon.social/api/v1/instance/peers |\
jq 'any(.[]; . == "onion.social")'

returned false. As I thought, moderated servers aren’t in the peers list.

The upshot of all of this is that Fosstodon hasn’t been defederated, at least not by any of the large instances I checked. So unless another crisis comes up, I’ll stay.

  1. The examples given from here on will be for mastodon.social. You can switch URLs to any instance to see similar information. ↩

The heat of the moment

May 23, 2025 at 8:54 AM by Dr. Drang

As part of a flurry of posts on Daring Fireball yesterday afternoon, John Gruber said something that really seemed directed at me. It was a sort of throwaway comment in the post about Apple Stores:

And the big partners, like CompUSA, absolutely sucked at showcasing the Mac. Their demo machines were frequently broken.

I was the reason one of those demo machines was frequently broken.

It was in the early 2000s, and there was a CompUSA near my office that I visited frequently. This was during my eight-year Mac hiatus, but I was still interested in what Apple was doing. I kept tabs on the state of the hardware and on OS X to see if and when it was time for me to return from the Linux wilderness.

One of the Macs my local CompUSA had out on display in its Apple ghetto was a G4 Cube. I had read somewhere that they ran hot, so one day I put my hand a few inches over the Cube to feel the convective flow rising out of the top grille. The machine immediately overheated and shut down to protect itself.

Let me be clear (as clear as the Cube). I didn’t put my hand on the grille—it was at least 3–4 inches above it. But that was enough. And from that day forward, every time I passed that Cube, I put my hand above it and caused a crash.

So in my own way, I was part of the success of Apple Stores.

A jarring font substitution

May 21, 2025 at 12:28 PM by Dr. Drang

I opened this article in Apple News+ this morning to see if, as I suspected, it was Long COVID that kept Kristaps Porziņģis from playing as much as usual during the NBA playoffs. It wasn’t, but before I even got into the story, I was taken aback by the headline.

Headline in Apple News

Not by the words in the headline, but by those two accented letters in his last name. They’re obviously not the same font as the rest of the letters—they’re not even seriffed. And whatever font they are (Helvetica, I believe), that font has a distinctly larger x-height than the font used in the rest of the headline, making the difference even more prominent.

I went to The Athletic’s website to look at the article there, and saw basically the same thing, but here at least, the two accented characters were seriffed. Still looked bigger than the surrounding text, though.

Headline on Athletic website

Here, the accented characters are almost certainly Georgia.

This unfortunate jump in font design and size is the result of font substitution. Whatever font The Athletic is using for its headline does not have those two characters—an n-cedilla and a g-cedilla—which are part of the Latvian alphabet (Porziņģis is Latvian). On Apple News+, the fallback font is Helvetica; on the web, it’s Georgia.

To see what font is being used for the rest of the headline, I went to WhatTheFont and uploaded an image of the headline with the accented characters removed. Many of its results were for some form of Cheltenham, which makes sense for two reasons:

  1. The New York Times, The Athletic’s parent, uses a Cheltenham variant for its headlines.

  2. The page source of the article includes a handful of references to “nyt-cheltenham” in its <style> section, like this:

    .article-content-container > h1, h2, h3, h4, h5, h6 {
  line-height: 1.2;
  font-family: nyt-cheltenham, georgia, 'times new roman', times;
}

    Which also explains why the substitute font on the web is Georgia.

So the NYT didn’t pay enough to get a full set of characters and Porziņģis’s name looks goofy in headlines. I’d say this was an obvious conspiracy by a bunch of Knick fans, but Porziņģis used to play for the Knicks.

Older posts