How big is Debian?

1. blog

Now this was quite a tease! For those who haven't seen it, I encourage you to check it out, it has a nice photo of a Debian t-shirt I did not know about, to quote the Fine Article:

Today, when going through a box of old T-shirts, I found the shirt I was looking for to bring to the occasion: [...]

For the benefit of people who read this using a non-image-displaying browser or RSS client, they are respectively:

   10 years
  100 countries
 1000 maintainers
10000 packages

and

        1 project
       10 architectures
      100 countries
     1000 maintainers
    10000 packages
   100000 bugs fixed
  1000000 installations
 10000000 users
100000000 lines of code

20 years ago we celebrated eating grilled meat at J0rd1’s house. This year, we had vegan tostadas in the menu. And maybe we are no longer that young, but we are still very proud and happy of our project!

Now… How would numbers line up today for Debian, 20 years later? Have we managed to get the “bugs fixed” line increase by a factor of 10? Quite probably, the lines of code we also have, and I can only guess the number of users and installations, which was already just a wild guess back then, might have multiplied by over 10, at least if we count indirect users and installs as well…

Now I don't know about you, but I really expected someone to come up with an answer to this, directly on Debian Planet! I have patiently waited for such an answer but enough is enough, I'm a Debian member, surely I can cull all of this together. So, low and behold, here are the actual numbers from 2023!

• 1 project: unchanged, although we could count 129 derivatives in the current census

• ~10 architectures: number almost unchanged, but the actual architectures are of course different (woody released with i386, m68k, Alpha, SPARC, PowerPC, ARM, IA-64, hppa, mips, s390; while bookworm released with actually 9 supported architectures instead of 10: i386, amd64, aarch64, armel, armhf, mipsel, mips64el, ppc64el, s390x)

• ~100 countries: actually 63 now, but I suspect we were generously rounding up last time as well (extracted with ldapsearch -b ou=users,dc=debian,dc=org -D uid=anarcat,ou=users,dc=debian,dc=org -ZZ -vLxW '(c=*)' c | grep ^c: | sort | uniq -c | sort -n | wc -l on coccia)

• ~1000 maintainers: amazingly, almost unchanged (according to the last DPL vote, there were 831 DDs in 2003 and 996 in the last vote)

• 35000 packages: that number obviously increased quite a bit, but according to sources.debian.org, woody released with 5580 source packages and bookworm with 34782 source packages and according to UDD, there are actually 200k+ binary packages ( SELECT COUNT(DISTINCT package) FROM all_packages; => 211151)

• 1 000 000+ (OVER ONE MILLION!) bugs fixed! now that number grew by a whole order of magnitude, incredibly (934809 done, 16 fixed, 7595 forwarded, 82492 pending, 938 pending-fixed, according to UDD again, SELECT COUNT(id),status FROM all_bugs GROUP BY status;)

• ~1 000 000 installations (?): that one is hard to call. popcon has 225419 recorded installs, but it is likely an underestimate - hard to count

• how many users? even harder, we were claiming ten million users then, how many now? how can we even begin to tell, with Debian running on the space station?

• 1 000 000 000+ (OVER ONE BILLION!) lines of code: that, interestingly, has also grown by an order of magnitude, from 100M to 1B lines of code, again according to sources.debian.org, woody shipped with 143M lines of codes and bookworm with 1.3 billion lines of code

So it doesn't line up as nicely, but it looks something like this:

         1 project
        10 architectures
        30 years
       100 countries (actually 63, but we'd like to have yours!)
      1000 maintainers (yep, still there!)
     35000 packages
    211000 *binary* packages
   1000000 bugs fixed
1000000000 lines of code
 uncounted installations and users, we don't track you

So maybe the the more accurate, rounding to the nearest logarithm, would look something like:

         1 project
        10 architectures
       100 countries (actually 63, but we'd like to have yours!)
      1000 maintainers (yep, still there!)
    100000 packages
   1000000 bugs fixed
1000000000 lines of code
 uncounted installations and users, we don't track you

I really like how the "packages" and "bugs fixed" still have an order of magnitude between them there, but that the "bugs fixed" vs "lines of code" have an extra order of magnitude, that is we have fixed ten times less bugs per line of code since we last did this count, 20 years ago.

Also, I am tempted to put 100 years in there, but that would be rounding up too much. Let's give it another 30 years first.

Hopefully, some real scientist is going to balk at this crude methodology and come up with some more interesting numbers for the next t-shirt. Otherwise I'm available for bar mitzvahs and children parties.