[rescue] Sun 3/80 Restoration

[Bill Dorsey]

Hi David,

On 10/6/19 1:39 PM, David Brownlee wrote:
> The 3/80 is an excellent SunOS 4.1.x platform - best of luck finding
> 4.1.1_U1 (IIRC) source :)
>
> On the NetBSD front - ssh is a particular issue as it keeps ramping up
> the computation effort for keys and connections over time as machines
> get faster. You might find it faster with a different ciper - "ssh -Q
> cipher" will list the available (or use rsh :-p)
Not sure if it's the fastest cipher to compute on the 3/80, but before 
giving up on NetBSD 8.0, I tried configuring SSH to use RSA with the 
smallest keysize supported -- 2048 bits IIRC.B  That was the 
configuration that took something like 45 seconds to log in with. For my 
purposes, 1024 bits is probably secure enough but I didn't take the time 
to recompile sshd with a reduced minimum on RSA keysize.B  Would that 
even work with current versions of ssh?
> I'm curious how much slower NetBSD would be compared to SunOS 4.1.1 in
> general use. Apart from gcc (which seems to be on a holy mission
> target memory use as factorial of version number) I would hope
> everything would be usable...

I've done some informal benchmarking between SunOS 4.1.1's cc, gcc-1.42, 
gcc-2.5.8, gcc-2.6.3 and settled on gcc-2.5.8 as a nice compromise for 
the 68k machines.B  2.6.3 was significantly slower at compilation while 
not producing any improvement in code size or execution size.B  Both 
2.5.8 and 2.6.3 produce better code than Sun's cc and 2.5.8 was actually 
faster than Sun's cc when using the optimizer although slower without 
the optimizer.B  I ruled out gcc-1.42 because it doesn't support C89 and 
because it produced worse code than Sun's cc.B  Getting glibc to work 
with 2.5.8 is an ongoing project though I don't think it will require 
much more effort.B  I'm not done experimenting with different compilers, 
but things are working well enough that this has fallen off my priority 
list.

I have had nothing but good experiences with NetBSD over the years I've 
been using it (since 1.4.0).B  However, the fact remains that most 
software has become more and more bloated with new features that some 
people consider desirable.B  This has a negative impact on older machines 
with (much) slower CPUs, (much) less memory and that few people still 
care about anyway.B  gcc, vim, and ssh are illustrative where the older 
versions generally run much faster on the older hardware than the newer 
versions.

- Bill