mathefuchs / al-for-sat-solver-benchmarking Goto Github PK

• clean experimental data to only contain relevant database from GBD
• choose a platform, e.g.,
  • GitHub
  • KIT's research-data service
  • some third-party provider like figshare (the latter has a dedicated space for TACAS, but might only be for validated artifacts)
  • rationale: GBD repo is public, but might be moved or force-pushed
• extend README of experimental data by describing structure of (most important) files (if this is not too much work); e.g., see the README ("Liesmich") of Markus' and my k-portfolio paper
• link experimental data from current (= code/text) repo's README

The steps described in README to address issue #2 are partly Linux-specific (shell script, some of the packages in the conda environment).

Some permissive license (like MIT) for the code would be nice. However, we should make clear this license does not apply to the whole repo (we probably cannot simply re-license files of the LNCS Latex template).

Eases reproducibility.

I need to clarify if I need to put in some acknowledgment or not.

If appropriate:

• add hints for developers working with the code base
• add a small demo of using the code, apart from running full pipeline (might be hard to find for this particular project)

E.g., see README of https://github.com/Jakob-Bach/Small-Portfolios.

While CC 4.0 was a good choice for the data repo, licensing software under CC is discouraged. I propose reverting commit c5dde33, so we use MIT again. Even if we kept the CC license, we would have the problem that we should not re-license the .cls and .bst file from the Springer template (our old MIT license had a disclaimer excluding these files).

... in the README. See https://github.com/Jakob-Bach/Small-Portfolios and https://github.com/Jakob-Bach/Constrained-Filter-Feature-Selection as examples.