(Key `MetrologyVocabulary` in https://gitlab.com/sosy-lab/admin/tex/-/blob/master/bib/sw.bib)
- repeat: same team with same technical means does the experiment again (check that experiment setup is deterministic and can be replicated by others)
- replicate: different team with similar technical means (use same implementation of algorithm; check that implementation works and reported data are valid)
- reproduce: different team with different technical means (reimplementation; check that idea works and is not flawed)
If this is not done then any change done by a user on the other OS
will cause "ghost changes", i.e., lots of lines changed when looking
at "svn diff" but they are only due to new line breaks, and it is
impossible to track real changes.
### Directory Structure
```
.
|-- documents # folder to keep the documents - submissions, reviews, etc.
|-- paper # folder for the paper source
|-- talk # folder for the sources of talks
```
### Process
- apply a tag on every major `stage` of the paper such as paper submission, proceedings, etc.
Example tag format: `<event>-<YY>-<submission, proceedings, talk>`
- create a directory named `documents` which will contain documents (used for interaction with others) related to the paper, e.g., submitted version, reviews, response, copyright, camera ready version.
Example format for doc name: `<YY>-<event>-<kind>`
- Branches and merges should typically not be used in order to facilitate change reviews. Merge commits can hide accidental changes because git by default does not show a full diff!
To make `git pull` always use rebase, do the following:
```
git config branch.master.rebase true
git config core.autosetuprebase always
```
## An advise that cannot be given often enough
> I think there is no substitute to a careful reading, in one sitting
> from beginning to end, in order to catch inconsistencies (mark them
> on paper; fix them later). Just searching for words or patterns
> within emacs often introduces more problems than it solves ...