David Brown([EMAIL PROTECTED])@Sun, Mar 23, 2008 at 11:51:19AM -0700:
> On Sun, Mar 23, 2008 at 10:04:06AM -0700, Carl Lowenstein wrote:
> 
> >It is a long-standing tradition that real computer users don't
> >need warm fuzzy encouraging status messages.  Programs that work
> >just do so silently, saving their breath to complain if
> >something went wrong.
> 
> Which, as far as I can tell, is completely untrue.  If something
> doesn't print out anything, they wonder if it's broken or if
> something went wrong.
> 

In the specific context of printing to STDOUT, I think this is
common in older utilities.  I wish it were more common.  The
convention of using the STD* file descriptors for their STD*
purposes is a very good one.  Need to spew other info?  Open
another descriptor and send it to the logger or someother file.

> Linus just recently changed the status threshold on git to be
> even shorter.  The old behavior was the if something was taking
> longer than 2 seconds, it would then start printing out progress.
> That's now 1 second.  Of course, git is divided up nicely between
> the plumbing programs, which are usually called by other scripts,
> and the porcelain which are not.  Even then, the porcelains
> usually have --quiet options.
 
Tools that have "--quiet" options are better than ones that don't,
but it still strikes me as a bandaid solution.  If I wan't verbose
output to STDERR, I'd rather ask for it and redirect, or do
whatever else I want with a utility that follows the formerly
widely respected conventions.

Wade Curry
syntaxman


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to