pacopy goes GPL

It looks like another outbreak of ‘clarify the licensing of the examples’ #445 arose overnight; following pygmsh #446, pacopy went from MIT to GPL, under

• commit ‘add setup.cfg’ https://github.com/nschloe/pacopy/commit/31369a497ea845e138ff30c53dcffa049b97c8f4#diff-9879d6db96fd29134fc802214163b95a
• PR ’modernize’ nschloe/pacopy#16.

While we could just do as for pygmsh #446 and explicitly GPL exx 23 (Bratu–Gelfand, using pacopy.euler_newton) and 27 (steady Navier–Stokes, using pacopy.natural), I’m inclined to suggest a different path.

I assume that the relicensing is not retrospective, so we could pin to the last MIT version. There has been very little development on pacopy since (my) addition of milestones in April 2019 in nschloe/pygmsh#5, which is essential to the practical application of pacopy.natural to nonlinear problems.

i had also recently already been thinking anyway of writing a new independent routines for natural parameter and false-transient continuation in the context of JFNK #442 #449 since I don’t think pacopy.natural is compatible with JFNK, running the Newton iteration itself. In simple cases like ex10, it’s enough to use a simple for-loop in the natural parameter (amplitude of boundary deflexion in that case). The adaptive natural parameter continuation of pacopy.natural in ex27 is nice though, so I’ll probably copy that from the last MIT version. Were I to write my own continuation routine, it would almost certaintly be an Iterator, avoiding the need for callbacks.

[kinnala]

Tangentially related: should be build a library of snippets that have only the relevant parts of interfacing to 3rd party libraries?

[gdmcbain]

the last MIT version.

That’s 0.1.2; 0.1.3 is GPLv3+.