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

[Frederik Ramm]

Hi,

Anthony wrote:
> So do I...  I just think it's unrealistic to expect it every time.

It is also unrealistic to expect good OSM data edits every time. Still 
it's good if people try, and good if the software helps them with it.

> IMO the job of the software should be to make sure the person knows
> the software has the ability to use comments.  Not to make it
> difficult for them not to use comments.

The software should also try and make people understand what comments 
are good for, i.e. why it is good to enter one, and perhaps give some 
information on what constitutes a "good" comment if the user is 
interested. Just saying "and by the way you can enter a comment here if 
you feel like it" is not enough.

> If we decide as a community that all edits *must* have comments, of a
> certain length, (and we shouldn't), then that should be enforced in
> the API.

I think there is room for things not enforced by the API but by an 
editor. Also, if we leave out "enforcing" and switch to "strongly 
suggesting" - which is exactly what my change did, if you remember, I 
removed JOSM's enforcing the 3 characters and replaced it by a nagging 
if you enter less than 10 -, then we leave API terrain altogether 
because the API has no way to do that.

> OTOH, if we're trying to get people to make comments that explain
> something that isn't evident from the edit itself, then 1) that's
> clearly unrealistic; 

I disagree.

> and 2) the changes to the software aren't really
> geared to that anyway (as no technical rules really can be).

You are right if you say there is no metric to measure the usefulness of 
a comment in software.

However, I have just randomly selected 100 comments of less than 10 
characters from the current end of changesets, and 100 comments of more 
than 10 characters, and there was a very noticeable correlation; in my 
non-representative sample, I found about 70 of 100 long comments useful, 
and I found about 10 of 100 short comments useful.

So while the software cannot *ensure* that people place a meaningful 
comment, it can certainly help with that by reminding the user if it 
seems likely that his comment is one of the "10 of 100" rather than one 
of the "70 of 100".

Bye
Frederik