[rescue] Re: [SunHELP] Solaris 9, First thoughts

[Patrick Giagnocavo]

On Sat, Jan 19, 2002 at 09:08:19PM -0500, Steve Sandau wrote:
> > > The problem is people code slow, sloppy, ineficcient code because it runs fine on
> > > the typical
> > > developer machine (which is almost always better/faster than the typical users
> > > machine).
> 
> We do the same type of thing by giving the developers decent machines,
> but also giving them a "test" machine that's a duplicate of their
> customers' machines. Easier for us because a customer's machine is
> fairly predictable (Navy office). "One" could do this by forcing
> developers to troubleshoot/test their work on a normal-to-slow machine.
> I bet this is something that *some* developers *don't* have to do... ;)

One thing that has been interesting to me is that the performance of
some of the "free" OSes on older hardware is actually *increasing*.
For instance, the revamped code between Linux 2.0 and 2.2 kernels
meant faster filesystem performance on trailing edge hardware.

And on my favorite OS, OpenBSD, I now find an IPC almost useful
whereas on 2.7 it was a dog.  They revamped the fs code and got a
substantial increase in performance, and other parts seem faster too.

And on a P75 doing ipsec tests I found an ssh session could maintain
over 100KBytes/sec in encrypted traffic, which would be enough for a
typical fractional T1 drop for ipsec use.

./patrick