On Jan 15, 2004, at 8:02 PM, Chuck Yerkes wrote:
> Just discussing elsewhere that I felt it was a waste of effort,
> 98% of the time, to spend efforts to keep utterly useless machines
> running.  While a VAX 750 holds some nostalgia for some people,
> to quote an Ikea commercial:  "Get over it, it's just a lamp."

This is a very valid reason to keep old slow hardware running.
On current processors, it can be very difficult to see that a
change resulted in sub-optimal behavior.  Since they are so
fast any inefficiencies may be lost in the noise.

For instance, when we moved the rc.d framework it was noticed
it was very slow on the less-powerful VAXes and 68ks.  Changes
were made to the shell and kernel to speed up forking and execing
resulting in a faster system *for all ports*.

> When developer time is spent getting a 2MHz machine to run well
> when it would be spent better getting current machines and new
> features in, then obscurity takes away from the project.  As neat
> as it would be to run BSD on my Motorola model 6300 Minicomputer
> (perhaps an S100 backplane, a 68030, an 80MB MFM disk and 1-2MB
> RAM and some wierd motorola unix that served 16 people in my old
> office until we replaced it with a 386/25 running Linux^H^H^H^H^HXenix
> that ran 200 times faster), any time spent to make that work would
> be taken away from useful projects.

Algorithms must be written more carefully to run quickly on a 20MHz
machine than a 3000MHz machine.  This machines also have much smaller
limits on memory and disk amounts.  By keeping the old machines running
they provide back-pressure on bloat.

> Faster file systems, logical volume management, ACLs, user-invisible
> clustering over a network, AFS ports - all more worthwhile that pushing
> off dust.

Not true.  Both have their places in development.

-- 
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this 
message.