My interest in GNU/Linux dates back to 1994 when a colleague, Nick Sheen, came into the post-graduate office with a CD-ROM of Slackware's Linux distribution. He seemed very keen on it, and being a physicist had it on his machine, so he put an installation on mine. I can't say I took to it immediately. Compared to certain other operating systems it seemed a little obscure, albeit, as Nick pointed out at great length, totally logical and coherent.
I kept the installation, dabbling with it every now and again, for a number of years. At the time it seemed unnecessarily complicated and difficult to do day to day tasks, such as write a letter, make up an illustration and get it into a document. However all this was about to change when I came to print out my Ph.D. thesis.
My thesis was about 120 pages of cross referenced text, diagrams and equations. The flagship wordprocessing package from a leading commercial vendor of software was simply inadequate for the task. It was bug-ridden, slow, difficult to use because of the reliance on mouse operation. To cap it all the printed version inserted a page break after every cross-reference, making the thesis about 300 pages long. The word-processor in question had proved it's worth writing letters, and the odd paper (although equations were difficult, and looked bad), but was simply not heavy duty enough to handle anything more substantial.
The only thing to turn to for future work was LaTeX/TeX running on top of GNU/Linux. It is nice and reliable, brilliant at cross-referencing, producing good looking papers, and equations. After some fiddling about postscript files are easy to get into documents, and look great. It was a steep
learning curve, but once enough had been learned to get a basic document to print out properly it actually becomes easier to use than some point and click abomination.
Obviously this change forced me to use GNU/Linux a lot more. I was still dual booting at the time, rather ridiculously to use GNU/Linux only for document processing. I was still using the leading branded spreadsheet, and statistical package to do most analyses. I managed to get an open source replacement
for the spreadsheet, but still couldn't do anything more elaborate than basic data manipulation with it. Then my colleague from Leeds,
Robert Aykroyd, introduced me to
R. The culture change between graphical spreadsheet to fully blown statistical language was no real problem as many academic statisticians use S-plus, which is merely another implementation of the same language specification.
R was the killer application for GNU/Linux as far as I was concerned. Superficially it is hard to use, with a somewhat unfriendly command prompt. But as with Unix stuff generally a little knowledge goes an awfully long way, and I soon ditched the graphical spreadsheet in its favor.
The only really difficult replacement was EndNote, which is a bibliographical database manager, available for non-*nix type systems. As a piece of software it is excellent, and for a couple of years I ran it from GNU/Linux via Wine, but latterly have replaced it with
Pybliographer by Frederic Gobry, which does pretty much the same thing.
These days I use GNU/Linux on my machine in the form of a
Departmental supported version of
Ubuntu. Other non-unix type operating systems are now relegated to recreational usage where their deficiencies are of no real concern.