meu entulho querido

notes

There's a problem with using Git concepts for everything

We’ve seeing a surge in applications that use Git to store other things than code, or that are based on Git concepts and so enable “forking, merging and distributed collaboration” for things like blogs, recipes, literature, music composition, normal files in a filesystem, databases.

The problem with all this is they will either:

  1. assume the user will commit manually and expect that commit to be composed by a set of meaningful changes, and the commiter will also add a message to the commit, describing that set of meaningful, related changes; or
  2. try to make the committing process automatic and hide it from the user, so will producing meaningless commits, based on random changes in many different files (it’s not “files” if we are talking about a recipe or rows in a table, but let’s say “files” for the sake of clarity) that will probably not be related and not reduceable to a meaningful commit message, or maybe the commit will contain only the changes to a single file, and its commit message would be equivalent to "updated ".

Programmers, when using Git, think in Git, i.e., they work with version control in their minds. They try hard to commit together only sets of meaningful and related changes, even when they happen to make unrelated changes in the meantime, and that’s why there are commands like git add -p and many others.

Normal people, to whom many of these git-based tools are intended to (and even programmers when out of their code-world), are much less prone to think in Git, and that’s why another kind of abstraction for fork-merge-collaborate in non-code environments must be used.