On 10/16/07, Stacey Jonathan Marshall - Solaris RPE
<stacey.marshall at sun.com> wrote:
> Chad Mynhier wrote:
> > On 10/15/07, Joseph Kowalski <jek3 at sun.com> wrote:
> >
> >> Seems good.  Even great.
> >>
> >> However, a bit of effort has been made to provide "more user friendly"
> >> output
> >> for many utilities.  For example, df(1M):
> >>
> >>
> >>      -h                          Like -k, except that  sizes  are
> >>                                  in a more human readable format.
> >>
> >
> > My original thought was to provide -u (microsecond) and -n
> > (nanosecond) options to allow users to specify the resolution.  For
> > these I would have left the default resolution at milliseconds for
> > backwards-compatibility.  That was the closest I came to the -h/-k
> > equivalent.
> >
> That's makes sense to me, so why were the options dropped from the
> consideration?

We saw a parallel to the case where ping -s output was changed to
provide microsecond resolution.  Millisecond resolution no longer
really makes sense for modern processors, nor for microstate
accounting, in much the same way that millisecond resolution no longer
made sense on modern networks.  It makes more sense to go ahead and
change the default.  If we accept that proposition, then it also makes
more sense to go ahead and change the default to be the resolution
that's available rather than set it to the intermediate resolution of
microseconds.

> > But I think you're actually asking for output like this:
> >
> That's what I thought but only if '-h' had been provided on the command
> line.

Yes, there should be a -h on that command line, like this:

# ./ptime -h -m -p `pgrep syslogd`

> > real    49h06m53s
> > user    35.72ms
> > sys     34.10ms
> > trap    154.1us
> > tflt    0
> > dflt    0
> > kflt    0
> > lock    343h48m03s
> > slp     245h34m21s
> > lat     65.53ms
> > stop    43.64us
> > #

Thanks,
Chad

Reply via email to