Sunday, February 21st, 2021

User-agent statistics
=====================

It's been more than a week since I asked the question "Is anybody out
there?"[1] and started to collect user-agent strings in (My Little)
User-Agent Database[2]. Now I consider my little experiment to be
mostly concluded as the most traffic happened during the first four
or five days and since then not much happens. 

Here are some stats.

I collected 1195 different user-agent strings. As I didn't prevent
bots from visiting the page, more than 60 % of these agents were bots
of various kinds. This itself speaks about what the modern
world-wide-web is - real people don't even count for one-half of the
traffic. I removed from the remaining 40 %, what I consider to be
utilities (curl, wget, netcat, etc.), not web browsers per se. The
remainder was 430 different strings. Maybe there were some from
non-human and/or non-browser sources, but if so, then they were well
camouflaged.

# Operating systems

+----------------------+---------+--------+
| O.S.                 | Strings |  Hits  |
+----------------------+---------+--------+
| Linux                |     117 |    267 |
| Android              |      83 |    135 |
| Windows              |      59 |    131 |
| Mac                  |      47 |    104 |
| iOS                  |      22 |     61 |
| BSD                  |       9 |     12 |
| Symbian              |       8 |      8 |
| Amiga                |       7 |     10 |
| IRIX                 |       6 |     10 |
| Nintendo             |       6 |      7 |
| Haiku                |       5 |     18 |
| SMART-TV             |       4 |      5 |
| Sailfish OS          |       3 |      8 |
| SunOS/Solaris        |       3 |      5 |
| ChromeOS             |       3 |      3 |
| Plan 9               |       2 |      7 |
| Palm OS              |       2 |      3 |
| Windows CE           |       2 |      2 |
| RISC OS              |       2 |      2 |
| Playstation Portable |       1 |      3 |
| Playstation 3        |       1 |      1 |
| Playstation 4        |       1 |      1 |
| OS/2 WARP            |       1 |      1 |
| DOS                  |       1 |      1 |
| Serenity OS          |       1 |      1 |
| Helen OS             |       1 |      1 |
| Blackberry 10        |       1 |      1 |
+----------------------+---------+--------+

# Linux by CPU architecture

+----------------------+---------+--------+
|  Architecture        | Strings |  Hits  |
+----------------------+---------+--------+
|  x86_64              |      79 |   216  |
|  i686                |      10 |    14  |
|  arm                 |       9 |    11  |
|  aarch64             |       3 |     5  |
|  ppc64le             |       3 |     4  |
|  ia64                |       1 |     1  |
+----------------------+---------+--------+

# Mac by CPU architecture

+----------------------+---------+--------+
|  Architecture        | Strings |  Hits  |
+----------------------+---------+--------+
|  Intel               |      30 |    76  |
|  PowerPC             |      17 |    28  |
+----------------------+---------+--------+

Some user-agent strings couldn't be assigned to a particular OS or
architecture, because none was reported - e.g Dillo or Lynx say don't
say a thing about what hardware and system are they running on. 

Yes, these data/statistics are not accurate. I wanted it to be "This
is what I use every day, let's give him a record in the database."
and soon it turned to be "This is what I dug of in the deeps of my
computer collection, let's give him a record in the database.",
which are two completely different things. But even so, there are
some interesting points, which are in fact more obvious this way:

1. People talk about many things, but they don't use them to actually
browse the web. For example, from what I usually read on Mastodon or
the phlogosphere, I thought that there are many active BSD users. But
when I sum all BSD systems found in the database, it's just nine
different strings and twelve hits. Yes, it places BSD in sixth place,
just behind Linux, Android, Windows, Mac, and iOS, but almost just as
many people found somewhere their old Nokia phone with Symbian (OS
dead since 2010). Not exactly widespread use.

2. There is no ongoing ARM-on-desktop revolution - not yet. The same
stuff as with BSD. Everybody has an ARM/AArch64 based SBC, people are
experimenting with it, talking about it, blogging about it, but not
replacing their PCs and Macs with it. And not even a single person
with Apple Silicon arrived. The x86/x64 hardware is still vastly 
dominant outside the mobile world.

3. There is someone out there with Itanium and a web browser on it!

As a result of my experiment, I can answer my original question: Yes,
there still is somebody connected online using non-mainstream devices.
Not many of them, but not zero. And I'm glad they arrived, especially
those with Amiga, Haiku, OS/2 Warp, or the single brave one with
Arachne browser on MS-DOS. Cheers to you all!

(Written using Geany in IceWM on RiscySlack/ppc64le - my daily-driver
hardware and software.)

[1] gopher://i-logout.cz/0/phlog/posts/2021-02-13_my_little_user_agent_database.txt
[2] http://uadb.1-2-8.net