Hi,
On Tue, Oct 07, 2008 at 02:43:10PM +0100, Eric Kow wrote:
> > As you very well put it ("I'd like"), this is a personal preference, not
> > necessarily a good user interface design choice.
>
> For background to this decision, see
> * http://lists.osuosl.org/pipermail/darcs-devel/2008-January/007011.html
> * http://bugs.darcs.net/issue896
>
> I would welcome any more background information behind this discussion.
> I had the impression that more people specifically asked for this
> behaviour before we implemented it. Is this something people feel
> strongly enough about to back out?
Here's my opinion, but be aware that I'm some really old-fashioned
hardcore unix user[1]:
- The size of the visible world is 80x24. Even on my big display, my
xterms are still 80x24.
- Therefore, the complaint in the first URL you mentioned has a point
(--help, or -? should produce *short* output). I'm not even sure
wether the advanced options should be printed by darcs at all
(this stuff belongs into the manpage).
- If the output has to be longer, the *user* should decide wether to use
a pager or not, i.e. don't pipe output to a pager from within the
program. The program (darcs, in this case) should not pipe anything
into a pager, if possible.
- Exception to the previous rule: if there are user interactions on
several independent entities (e.g. patches), launching a pager
from the program for every single entity is ok. For darcs, this
would include interactive selection of patches during a pull, and
darcs already does a nice job here, since you can choose wether
you want the current patch written to stdout or piped into a
pager.
- Usage output (be it via --help, -?, --gimmeaclue) should be written to
stdout, not to stderr -- AFAIK this isn't an issue for darcs, but I
hate it *so much* if I've to type something like
foo --help 2>&1 | more
for some `foo' tool that I mention it here ;-)
> Two other notes
> * to get help explicitly without a pager: darcs help | cat
> to get help explicitly *with* a pager: darcs help | less
> so as I understand it, this is a question of picking the right
> defaults
Well, for cherrypicking patches interactively, you *need* a pager
(which is ok), and it *has* to be launched by darcs (which is ok).
For one-shot output, darcs shouldn't even try to be smart. Just
dump the pile of characters to stdout.
> * one potential disadvantage of setting DARCS_PAGER to cat is that
> you lose the ability to explicitly request paging, for example,
> hitting 'p' in interactive mode to view the current patch through
> a pager
Exactly. So don't pipe --help output or anything else that isn't
interspersed with user input in interactive mode into a pager.
Ciao,
Kili
[1] Not really old-fashioned; I never used a real tty, my first
contact was via an ultra-modern vt220 clone. When they started to
deploy (hardware) X terminals at the campus, I was a little bit
upset about the fact that you had a huge 20 inch monitor with a
grey backround and only that tiny 80x24 working aread in the upper
left, so I crawled back to the room with the vt's. A year later
some friend told me how to write a proper .Xsession ;-)
--
No, I'm not afraid to publish my mail address on the web. I get so
much spam that a few hundered more each day won't make any noticeable
difference.
-- Artur Grabowski
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users
