Trial git-based task list

[Daniel Shahaf]

Ximin Luo wrote on Thu, Nov 10, 2016 at 18:55:00 +0000:
> Holger Levsen:
> > I'm sorry if this sounds dismissive, but this thread (and evaluation)
> > has shown me, that being decentralised is not a feature I desire in a
> > tracker, on the contrary, it seems that decentralised has downsides
> > making me wish for a centralized tracker which I can use with a
> > webbrowser.
> > 
> > (or someone needs to setup a webview for this tasks.git thing. fine with
> > me too.)
> > 
> > As I understand, this tasks.git needs me to review code to use it,
> > which… (here!) basically means "no". We have a zillion trackers in
> > Debian, why not use a packaged one, preferedly on a server.
> 
> I wouldn't generalise this to all decentralised trackers. The current
> setup is a massive hack for sure. But yes let's talk about it some
> more next meeting.

Would rather discuss it as much as possible on list so as to save time
during the meeting.

I think our choices are:

- Use taskwarrior with less (or no) code to review.

  Is there a way to minimise the amount of code one has to review?
  Right now, there is a custom merge driver, a custom 'task' wrapper,
  custom git hooks and custom taskwarrior hooks.  How many of those can
  we do away with?

  I think something like this might fly with people:

      % cat task.in
      #!/bin/sh
      set -e
      export TASKRC=/dev/null
      export TASKDATA=$(mktemp -d)
      taskdir=$PWD/taskdir
      cp $taskdir/*.data $TASKDATA/
      (set +e; task "$@"); rc=$?
      if [ $rc -eq 0 ]; then cp $TASKDATA/* $taskdir/; fi
      rm -rf $TASKDATA || true
      exit $rc

  It's short enough to review, and doesn't run any code from the
  repository by default (the 'task.in' file in git is not executable).
  It does nothing about resolving git conflicts, though.

- Use another distributed system that requires little (or no) code to review.

  Plain text files?  Anything else?

- Use a system with a server.