Re: Using VC for change descriptions

[Paul Smith]

On Thu, 2018-05-10 at 13:23 +0200, Thien-Thi Nguyen wrote:
> My understanding is that RMS is open to evolution of the
> ChangeLog format, but only if it is non-subtractive.  (FWIW,
> this is my personal stance, too, so i may be projecting here.)
> IOW, 1+"changed-entities" SHOULD NOT be dropped.

I don't think that's what RMS said exactly.  But maybe I'm projecting
as well.

I don't see how this position ("non-subtractive ChangeLog format")
moves us anywhere.  This is exactly what we have today: we can auto-
generate ChangeLog files from Git message logs.  There is a script in
gnulib which does it, and multiple projects already take advantage of
it.

The point being made is that requirements this forces on the format and
content of Git commit messages goes against what some project
maintainers want to use, and causes people to provide worse / less
useful commit messages than they otherwise would.

Unless what you're suggesting is that instead of ChangeLog files
formatted like this:

  * entity1: change to entity1
  * entity2: change to entity2
   ...
  * entityN: change to entityN

That a format something like this would be acceptable:

  description of changes in the commit

  * entity1
  * entity2
   ...
  * entity3

That is, the description comes first then a bare list of changed
entities, without explicit descriptions attached to each one.  This
ChangeLog format could be generated using a simple tool which extracted
the Git commit message then computed and appended the list of modified
entities for that commit.

I have no real objection to this, although I personally don't see much
advantage to it.