On Mon, Apr 04, 2011 at 11:40:43 +0100, Miles Gould wrote: > 1) I don't think I've ever used "darcs unrevert". > 2) I do, on the other hand, use "git stash" A LOT. It was one of the > features that caused my last team to adopt git. If it had been clear to > me that revert/unrevert could be used to implement "darcs stash", I'd > probably have used them for that purpose.
Thanks for the feedback! I ♥ git stash too and would probably use it if Darcs had such a feature. Aside from the conflicting principles we're juggling, we have the problem of trading off between user types. > 4) My least favourite thing about darcs is its chattiness. Removing > a "do you want to do this?" prompt is IMHO a win. I agree. I hate the unrevert prompt too and I applaud any thinking that would lead to it going away. I also think that these qualities are desirable * you can undo the revert operation * no need to anticipate wanting to undo that revert operation * no need to remember where you saved your revert work * simple retrieval mechanism * no need to remember special incantations (symmetry helps) So what I think we should really do is ask ourselves why the "this will make unrevert impossible" confirmation prompt exists, and how we can kill that prompt. We've seen four ideas so far: 1. status quo 2. revert -O 3. .bak files 4. git stash/unstash 5. _darcs/trash (Ganesh on IRC) I'm not entirely against #4; I could see how it's actually quite compatible with these goals. I'm little bit worried about the stash stack. Is it too sophisticated? Even for non-newbies, does it impose any extra cognitive/memory work along the lines of "which stash do I recover and how?" or would we all just tend to push/pop anyway? Does anybody have any other ideas? > 3) darcs' proliferation of undo commands is confusing to newbies. > Speaking of which, I note that unrevert gets only a cursory mention on > http://en.wikibooks.org/wiki/Understanding_Darcs/Undoing_mistakes Good. Thanks for pointing this out. My complaint is that we are all speaking from our expert perspectives which is not entirely healthy. Let's suppose we had the following back-of-the-envelop "personas" [1] to consider: * Jacek - heavy user; maintainer; expert sophistication Goals: keep repository history clean, well-structured; avoid cruft * Steve - regular user; programmer in team; medium/low sophistication Goals: be able to save things into version control and get them back older versions later on; get back to hacking as soon as possible! * Adam - casual user; hobby open source contributor; medium sophistication Goals: avoid fuss when submitting patches to FooBox open source project * Mike - newbie; programming company intern Goals: get started quickly, avoid frustration of dealing with incomprehensible new tool I'm making stuff up here (I admire Cooper-style ethnographic research used to get more informed versions of these), so they may be bogus. But assuming there is some truth to these, one problem is that the kind of user who is going to chime in on this list is more likely going to be a Jacek or maybe a Steve than an Adam or Mike. Miles has done a nice job of bringing Steve (his last team) and Mike (newbies confused by all the undo commands) into the picture. We need to hear a bit more about users from the other parts of the spectrum that we care about. Maybe I'm worrying over nothing. You give Adam and Mike git stash/unstash and they would love it! -- Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow> For a faster response, try +44 (0)1273 64 2905 or xmpp:[email protected] (Jabber or Google Talk only)
signature.asc
Description: Digital signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
