Time flies. On this day, 10 years ago, a certain someone sent in his first contribution to Debian in Debbugs#433007: --dry-run can mark a package manually installed (in real life). What follows is me babbling randomly about what lead to and happened after that first patch.
That wasn't my first contribution to open source: I implemented (more like copy-pasted) mercurial support in the VCS plugin in the editor I was using back in 2008: Geany – I am pretty sure my code is completely replaced by now, I just remain being named in THANKS, which is very nice considering I am not a user anymore. My contributions to apt were coded in vim(-nox) already.
It was the first time I put my patch under public scrutiny through – my
contribution to geanyvc was by private mail to the plugin maintainer –
and not by just anyone but by the venerable masters operating in a mailing list
I had started looking into apt code earlier and had even written some patches for me without actually believing that I would go as far as handing them in. Some got in anyhow later, like the first commit with my name dated May the 7th allowing codenames to be used in pinning which dates manpage changes as being written on the 4th. So then I really started with apt is lost to history by now, but today (a decade ago) I got serious: I joined IRC, the mailing list and commented the bugreport mentioned above. I even pushed my branch of random things I had done to apt to launchpad (which back then was hosting the bzr repository).
The response was overwhelming. The bugreport has no indication of it, but Michael jumped at me. I realized only later that he was the only remaining active team member in the C++ parts. Julian was mostly busy with Python at the time and Christian turned out to be Mr. L18n with duties all around Debian. The old guard had left as well as the old-old guard before them.
I got quickly entangled in everything. Michael made sure I got invited by Canonical to UDS-L in November of 2009 – 6 months after saying hi. I still can't really believe that 21y old me made his first-ever fly across the ocean to Dallas, Texas (USA) because some people on the internet invited him over. So there was I, standing in front of the airport with the slow realisation that while I had been busy being scared about the fly, the week and everything I never really had worried about how to get from the airport to the hotel. An inner monologue started: "You got this, you just need the name of the hotel and look for a taxi. You wrote the name down right? No? Okay, you can remember the name anyhow, right? Just say it and … why are you so silent? Say it! … Goddammit, you are …" – "David?" was interrupting my inner voice. Of all people in the world, I happened to meet Michael for the first time right in front of the airport. "Just as planned you meany inner voice", I was kidding myself after getting in a taxi with a few more people.
I meet so many people over the following days! It was kinda scary, very taxing for an introvert but also 100% fun. I also meet the project that would turn me from promising newbie contributor to APT developer via Google Summer of Code 2010: MultiArch. There was a session about it and this time around it should really happen. I was sitting in the back, hiding but listening closely. Thankfully nobody had called me out as I was scared: I can't remember who it was, but someone said that in dpkg MultiArch could be added in two weeks. Nobody had to say it, for me it was clear that this meant APT would be the blocker as that most definitely would not happen in two weeks. Not even months. More like years if at all. What was I to do? Cut my looses and run? Na, sunk cost fallacy be damned. I hadn't lost anything, I had learned and enjoyed plenty of things granted to me by supercow and that seemed like a good opportunity to give back.
But there was so much to do. The cache had to grow dynamically (remember "mmap ran out of room" and feel old),
commandline interfaces needed to be adapted, the resolver… oh my god, the resolver! And to top it all of
APT had no tests to speak of. So after the UDS I started tackling them all: My weekly reports for GSoC2010
provide a glimpse into the abyss but before and after lots happened still. Many of the decisions I made
back then are still powering APT. The shell scripting framework I wrote to be able to perform some automatic
testing of apt as I got quickly tired of manual testing consists as of today of 255 scripts run not only by
me but many CI services including autopkgtest. It probably prevented me from introducing thousands of regressions
over the years. Even through it grew into kind of a monster (2000+ lines of posix shellscript providing the test framework alone),
can be a bit slow (it can take more than the default 30min on salsa; for me locally it is about 3 minutes) and it has
a strange function naming convention (all lowercase no separator: e.g.
insertinstalledpackage). Nobody said
you can't make mistakes.
And I made them all: First bug caused by me. First regression with complains hitting d-devel. First security bug. It was always scary. It still is, especially as simple probability kicks in and the numbers increase combined with seemingly more hate generated on the internet: The last security bug had people identify me as purposefully malicious. All my contributions should be removed – reading that made me smile.
Lots and lots of things happened since my first patch. git tells me that 174+ people contributed to APT over the years. The top 5 of contributors of all time (as of today) list:
- 2904 commits by Michael Vogt (active mostly as wizard)
- 2647 commits by David Kalnischkies (active)
- 1304 commits by Arch Librarian (all retired, see note)
- 1008 commits by Julian Andres Klode (active)
- 641 commits by Christian Perrier (retired)
Note that "Arch Librarian" isn't a person, but a conversion artefact: Development started in 1998 in CVS which was later converted to arch (which eventually turned into bzr) and this CVS→arch conversion preserved the names of the initial team as CVS call signs in the commit messages only. Many of them belong hence to Jason Gunthorpe (jgg). Christians commits meanwhile are often times imports of po files for others, but there is still lots of work involved with this so that spot is well earned even if nowadays with git we have the possibility of attributing the translator not only in the changelog but also as author in the commit.
There is a huge gap after the top 5 with runner up Matt Zimmerman with 116 counted commits (but some Arch Librarian commits are his, too). And that gap for me to claim the throne isn't that small either, but I am working on it… 😉 I have also put enough distance between me and Julian that it will still take a while for him to catch up even if he is trying hard at the moment.
The next decade will be interesting: Various changes are queuing up in the master branch for a major break in ABI and API and a bunch of new stuff is still in the pipeline or on the drawing board. Some of these things I patched in all these years ago never made it into apt so far: I intend to change that this decade – you are supposed to have read this in "to the moon" style and erupt in a mighty cheer now so that you can't hear the following – time permitting, as so far this is all talk on my part.
The last year(s) had me not contribute as much as I would have liked due to – pardon my french – crazy shit I will hopefully be able to leave behind this (or at least next) year. I hadn't thought it would show that drastically in the stats, but looking back it is kinda obvious:
- In year 2009 David made 167 commits
- In year 2010 David made 395 commits
- In year 2011 David made 378 commits
- In year 2012 David made 274 commits
- In year 2013 David made 161 commits
- In year 2014 David made 352 commits
- In year 2015 David made 333 commits
- In year 2016 David made 381 commits
- In year 2017 David made 110 commits
- In year 2018 David made 78 commits
- In year 2019 David made 18 commits so far
Lets make that number great again this year as I finally applied and got approved as DD in 2016 (I didn't want to apply earlier) and decreasing contributions (completely unrelated but still) since then aren't a proper response! 😉
Also: I enjoyed the many UDSes, the DebConfs and other events I got to participate in in the last decade and hope there are many more yet to come!
tl;dr: Looking back at the last decade made me realize that a) I seem to have a high luck stat, b) too few people contribute to apt given that I remain the newest team member and c) I love working on apt for all the things which happened due to it. If only I could do that full-time like I did as part of summer of code…
P.S.: The series APT for … will return next week with a post I had promised months ago.