[josm-dev] Change to changeset comment handling, RfD

[Ævar Arnfjörð Bjarmason]

On Sat, Jul 31, 2010 at 19:23, Frederik Ramm <frederik at remote.org> wrote:
>   I have changed the way JOSM treats changeset comments. This is a
> politically sensitive topic (see recent discussion on talk list), so I'd
> value your input. I'm willing to revert the change (applied in r3399) if
> people think it is for the worse.

Maybe I shouldn't have said anything :)

> Some people do not want to enter changeset comments, and thus are unhappy if
> their editor forces them to.
>
> Some people want to enter changeset comments and want everyone else to do
> it, but they fear that forcing people to enter something will only create
> bitterness, and people will start entering useless comments like "some
> fixes", "..." and the like.
>
> Some people think that the editor must not give people an excuse not to
> enter a changeset ("but I didn't know it was required...") and thus want the
> editor to enforce that as far as possible (of course no editor can check
> whether what's entered makes sense).
>
> For a while now, JOSM has forced people to add a changeset comment that is
> at least three characters long, and refused to perform an upload otherwise.
> In addition, JOSM *always* loaded that last recently used changeset comment
> as a default, making it easy for people to erroneously continue using an old
> comment.

Since there's no way to turn off the dialog permanently that I can see
I fear that we'll simply get "..........." instead of "..." from the
people determined not to insert a changeset comment, but maybe not.
We'll see.

> I have now changed this in the following way:
>
> * The last recently used changeset description is only loaded as default if
> it was used less than 4 hours ago. Otherwise there is no default.

FWIW my main issues with the changeset dialog have been that it's too
easy when you're juggling multiple changesets to mistakenly use the
changeset summare you meant to use for changeset A for changeset B.
I've filed a lot of bugs related to this (none of which I can find
now, bad trac search..). There's a lot of little UI issues with that
feature.

> * If you try to upload a changeset with a comment of less than 10 (!)
> characters, there's a dialog that explains the importance of changeset
> comments, and asks the user to reconsider. The user may refuse to
> reconsider, in which case JOSM will proceed with the upload, even if the
> changeset comment is empty.

Does the .length() method measure bytes, or does it measure
characters. I looked at the changes over Japan and it seemed that
there were some useful comments that also fit within 10 characters.
CJK languages have a lot more information density per character than
we're used to here in Europe.

Perhaps just counting the bytes would be a better heuristic, if this
route is to be taken that is.

> I think that this strikes the right balance between trying to educate the
> user and still giving him the freedom to say no.
>
> I do have an user interface problem however. The dialog I'm displaying is a
> bit funny. It has the buttons "yes, reconsider" and "no, continue uploading"
> (because I *want* people to reconsider - thus the fat green check mark with
> that option). But now if you see the dialog and press ESC, it proceeds with
> uploading, which is undesirable. I don't know how to solve this. Any ideas?

I'm not familiar with the Java APIs in question, and maybe I'm too
used to webapps, but can this sort of thing be done by expanding a
textbox within the upload dialog itself, instead of having a new popup
window?

JOSM is sometimes a maze of popup windows, all alike :)

> As I said, if people think this makes things worse, we can always go back.
> It's a difficult thing. Nobody likes to be bossed around by software - but
> then again it's no good if the software makes it sound like everything is ok
> and later you get flamed on the mailing lists (by people like me) for not
> using changeset comments ;)

Phoning home is a bad thing, but it would be good to have some data
points on whether any of this nannying around is actually working. It
probably is, most users are probably trying to do the right thing, and
might not be aware of the importance of changeset comments until it's
pointed out to them.

It'd be nice to data mine the changeset comments for the whole planet
made with JOSM before and after this change. Maybe we'll get better
(measured by some simple metric, like # of words) and longer comments,
but maybe the comments will just be longer garbage.

It'd be good to know that, so that we know that features like these
aren't just pointlessly annoying users.