Faith in open source software tested to the limit.

I'm a big supporter of open source software, I've used linux as my main operating system for the last 10 years (if not more). At work, as a scientist, I've made use of the large catalogue of scientific software and tools on offer to Linux users. I've always been a staunch supporter of open source over propriety software, recommending people seek out the alternatives before parting with cash for mainstream solutions.

However, over the last few years I've been noticing an increasing problem that I think needs addressed before there is mainstream adoption of Linux as a desktop solution. Namely: Brilliant software ruined by less than optimal upgrade cycles. More regularly than I would like, I update my system and then find that software that I base my work (and hobby) life on no longer works.

Scilab

Scilab is a mathematical software package. I've used it since I was an undergraduate physics student (graduating in 2005) as an alternative to Matlab. It formed the basis of my PhD thesis work in to signal processing and analysis, and I even delved in to controlling hardware with it's SciCos modules. Now however, I need to find an alternative because, after updating my system, Scilab's push to a new underlying API has been at the expense of being able to plot data on intel hardware!

No longer can I read simple data in and plot graphs. This is the simplest use of Scilab, and one that formed the basis of many images in my thesis and papers. I could understand if it was obscure hardware that the graphical library didn't work with, but seriously, Intel!??!? 

Mendeley

Mendeley is a wonderful tool for sorting out academic papers and keeping a track of citations. In a lot of ways, I would still recommend it. However, today, while trying to make headway in writing a few articles, I have found a very serious bug that makes it all but useless for its task: Mendeley is unable to insert citations in to OpenOffice. 

The developers have known about this bug since April, other updates and bugfixes have been issues, but this seems to be less than urgent from their point of view. From an academic point of view, I am no longer able to keep a dynamically updated bibliography section at the bottom of my paper as I write it. I can't even edit the existing reference list.

JackD

Ah Jack, the godsend for Linux Audio, which worked brilliantly for a good number of ages before some bright spark decided that it should go the same was as Pulse Audio (ugh), and upgrade to JackD2. Without warning, the wonderful multitrack studio software studio Ardour is unable to export audio without disabling all audio plugins (EQ, reverb, compressers etc). Fine if you want to render dull, unmixed studio audio, but that then misses the whole point of what a Studio environment is for.

Why?

It seems that developers of high profile open source software want to be seen to be staying ahead of competition, and by making use of the free nature of the upgrade cycles afforded by open source, rush out these upgrades without any proper testing. "The audience will test it" seems to be the mantra, forgetting that there is a reason that the community should make use of Nightly Builds (new stuff, not tested), and Stable builds (more stable than nightly, but needs more testing), and Release Candidate (stable as fuck, as good as the existing sofware but with new and WORKING features). 

Unfortunately, we are unwittingly being upgraded to Stable builds rather than Release Candidates and these problems arise.

Right, I'm away to find alternatives, until they break too.