Picked up a battered Fortran reference book that someone had left behind in the math grad students' dingy basement. There is rationally no reason to keep books now that everything can be looked up on the internet. And if you get an error you're more likely to find a solution to that specific error message by searching Google Groups for people who've reported similar errors. Nevertheless, the book provides a sense of security that the internet doesn't. Better a material slab of flattened dead trees than ephemeral networks of pooled expertise.
Just spent five hours discovering that PAW wouldn't accept my code because I had used a slightly different variable list to the one that describled my ntuples. There being a few hundred variables on the list, I hadn't bothered to check the individual terms to see that they were the same. The graphs I've gotten after the online veto are barely less horrendous than the raw data. Still, a slight sense of satisfaction, which I had the pleasure of enjoying weekly when I was taking Computational Physics. Perhaps I'm just not getting the right level of challenges in my recent maths/physics courses? Complex anal is just so far above my head that there isn't even a chance for me to meaningfully confront the challenge. The Griffiths QM problems are challenging on my patience but not on my physical intuition or reasoning. In sharp contrast to the problems we got in introductory mechanics and electrodynamics.
The original point of this post was to complain about the lack of any introductory manual to PAW. Everyone complains about it, but then they just learn it the hard way, and after they've learnt it, they can't be arsed to write a tutorial themselves. To put it in economic terms, an introductory manual is a public good. Still, one can find great, free introductions to things like LaTeX and the various programming languages online. Perhaps it takes a critical mass of users for there to exist altruists who'll write an introductory text.