On Thu, Feb 22, 2018 at 09:00:45PM +0100, Lars Schneider wrote:
> > If it was only about a diff of UTF-16 files, I may suggest a patch.
> > I simply copy-paste it here for review, if someone thinks that it may
> > be useful, I can send it as a real patch/RFC.
> That's a nice idea but I see two potential problems:
> (1) Git hosting services (GitLab, BitBucket, GitHub, ...) would still
> show these files as binary. That's a huge problem for my users as
> they interact more with these services than the Git command line.
> That's the main reason why I implemented the "UTF-8 as canonical
> form" approach in my series.
I can't speak for the other services, but I can tell you that GitHub
would be pretty eager to enable such a feature if it existed.
I suspect most services providing human-readable diffs would want to do
the same. Though there are still cases where you'd see a binary patch
(e.g., format-patch in emails, or GitHub's .patch endpoint, since those
are meant to be applied and must contain the "real" data).