Re: Should we restore manually maintained ChangeLogs

[Paul Eggert]

On 03/09/2016 11:14 AM, Eli Zaretskii wrote:
> This community is about software development, not about historical 
> research. When I'm looking up a commit, I want accurate information 
> about it.

Like it or not, that is a form of historical research. One cannot escape 
the basic principles of history, and in particular one cannot insist 
that the historical record must be error-free.


> > There is a reasonable question about how much of our development effort
> > should be devoted to sprucing up ChangeLogs after they're committed. I
> > think this should be low priority, whereas as I understand it you would
> > prefer that we boost its priority. Neither side is advocating
> > untrustworthy ChangeLogs, or perfect ChangeLogs for that matter; it's
> > mainly a question of where to allocate our scarce development resources.
> I'm arguing that we shouldn't _need_ to allocate resources to it.

There is no free lunch here. There is a real cost to the old-fashioned 
approach of keeping commit messages as files in the repository. This 
cost is borne by every contributor, and the hassles of dealing with it 
was a primary motivation for Emacs (and other projects) moving away from 
that approach. Regardless of the approach taken, there is also a cost to 
sprucing up the historical record, a cost borne by the developers who do 
the sprucing-up. This sort of approach does work unless we devote real 
resources to it.