> Steven N. Hirsch wrote:
> >
> > Ingo, et al,
> >
> > Ok, I admit it. I'm a horse's rear-end.
> >
> > After many gripes about Midnight Commander being extremely slow under
> > 2.1.x kernels when stat-ing in large directories, I finally came to my
> > senses and realized that the target directory was actually a symlink to an
> > NFS volume :-(. The link is only 10Base-T, so of course it was fairly
> > slow.
> >
> > Many, many apologies. I'm not often this dense, but when I am I do it
> > well. Sorry for the bandwidth and effort wastage <groaning and ducking>.
>
> Let me get this straight: You mean that you had two different
> machines, one of which had NFS mounted that directory that went
> quickly on the other machine?
Yes, I had a complete mental block. I was laboring under the belief that
the /usr/man hierarchy on my test box was a _copy_ of the one on the
server (peer, actually).
>
> There are MANY MANY effects that can cause performance differences.
> In the future (anybody listening?) please make sure you really
> eliminate all other possibilities by really running the test at least
> three times on the same machine:
>
> case A
> case B
> case A
>
> to make sure that 1) there are no other factors involved. 2) that not
> a sudden jump in the network load or something influences your
> results. (3: 3 is too small a number)
No need to rub it in. I totally screwed up.
Steve