Rants from the Ballmer Peak rss feed

Small big mistakes

Open software is cool. Everybody can join in and tinker with the code. But sometimes the entrance threshold is too high and projects are kept under the radar for most people. I believe the Nim programming language by Andreas Rumpf & Contributors slightly tiptoes into the could-be-much-better-with-a-little-more-care. The disadvantage of Nim is that it is a programming language, and a really full fledged one which beats most commonly used ones in terms of flexibility and amazing features. Most potential contributors are intimidated by the initial complexity of such software (not everybody writes compilers for breakfast).

But even more pitiful is what I consider basic administration mistakes which discourage potential users. It is not fun when users have to report that the build instructions fail (wait a second, why are build instructions not part of the continuous integration tests and the issue is neglected for weeks or months?). It is even less funny that this is neglected because somebody has to manually upload files to the website, and starts to look like a bad joke when after updating the website, some pages are still missing due to human error, so another manual interaction is needed.

For a group of people capable of creating a programming language the only possible answer is lack of attention to releases. When programmers who don't like repeating themselves (sometimes called lazy but in a positive way) are confronted with a manual task, there are two possible paths: avoid the task or automate the hell out of it. In this case, Nim's visibility is hurt, as new users come and find something which doesn't work at first glance. Why should they keep bothering when there's a constant torrent of new shiny programming languages out there competing for everybody's attention? In more technical terms, it doesn't matter how many users have write access to the repository if updating the website still has a bus factor bottleneck.

Some time ago Babel, the official package manager for Nim was in a state of flux where some stuff did not work correctly, or required some prodding. I don't understand what is so hard about keeping the repository master branch compilable. Maybe I'm weird. But it get's better! Here's something to facepalm if you want to use Babel to install Aporia, the official Nim editor, linked from the Nim webpage:

$ babel install aporia
Downloading aporia into /tmp/babel/aporia using git...
Found tags...
blah
blah
more blah
and then…
FAILURE: Specified directory does not contain a .babel file.
$

So the secret handshake this time is to run babel install aporia#head. Totally obvious. The Aporia repository uses tags, but unfortunately the last one didn't have a .babel file with the package info, so the #head suffix fixes that. As if version numbers were expensive and you could not simply bump the version number and re tag the repository with the added file. On the bright side Aporia specifies this in the compiling section of the README, but it is a weird bright side to be, demanding all your potential users to jump through hoops when you can easily avoid that collective pain.

But it all makes you wonder: how many people did find Aporia listed in Babel's list command output and tried to install it? Hint: IRC logs show a few. How many didn't bother asking on IRC? We will never know… How many users did tried to compile Babel during the time it was unstable? Hopefully few. How many users did Nim lose to outdated github build instructions, since pretty much every other Nim software in active development requires the git version? These are all trivial things to solve really.

Oh, and how much do I hate people graciously giving commit access to github repositories. Let me explain: if this works for you it is only because you are a one man shop or don't really care about the project and are looking for somebody to replace you. Can you imagine a project like Webkit or the Linux kernel giving out write access to just about everybody? Not going to happen, that software has value, and random changes by random people are unlikely to increase value. Sorry, the pull request hack only works if your project has essentially no value to you, or if you think 100% of humans are good, something which is easily disputed. It's the reason we have captchas to give our opinion on many websites.

Unfortunately I'm preaching to the walls: in my professional life as a software developer I've seen even worse blunders in commercial software. You know, the one where people's heads roll when something goes wrong. So my theory is that most people doing open source software simply replicate what they have learned at work. Or worse, are too lazy to even do that minimum effort!

That's when <inception_bwaaaa> I step in to save the world of software </inception_bwaaaa>. Actually, I did read recently some hilarious articles by James Mickens (scroll to the bottom of that page, or find the extracts and direct links at the Programming is terrible blog which is also a nice read), so I thought to myself, hey, why not try and write something absurd while being semi serious about the issue? So here goes my paper in pedantic PDF form:

Hmm… maybe that wasn't such a good idea after all.

$ nim rst2html paper.rst
paper.rst(5, 1) Error: invalid directive: 'humour'


See the article index or browse articles by tags: programming , nim , bureaucracy .
Published on: 14/01/2014 08:02. Last update: 23/02/2014 20:12. rss feed
Copyright 2021 by Grzegorz Adam Hankiewicz.
Generated with ipsum genera. Look at the source code.