Re: the problem of rust :-(

[Valery Ushakov <uwe%stderr.spb.ru@localhost>]

On Thu, Jun 06, 2024 at 07:33:36 -0400, Brad Spencer wrote:

> When the VAX (32 bit system) came out, it took a while before the PDP-11
> (16 bit system) was replaced and I am sure that sales of the PDP-11
> continued for some time (and probably for DECs competition).  Indeed,
> even last year in 2023 a patch for BSD 2.11 for the PDP-11 appeared.
> However, one can't really argue that as the 32 bit systems became more
> common, software that would run on the 32 bit systems would often times
> not run on the 16 bit systems.  Some efforts were made to port the
> software but it was often quite a hard task.  I lived though the very
> end of this transition.

I started on PDP11, I jumped through quite some hoops to build various
programs for Xenix/286, so I do have some first hand experience.  The
16-bit vs 32-bit was a huge change.  2^16 is only 64K, and even with
separate instruction and data address spaces, that's not much.  E.g. I
never managed to get cnews to fit, so I had to run bnews on that Xenix
for usenet.

32-bit vs 64-bit is _nowhere near_ in comparison for most programs,
unless you are doing industrial scale OTP or some huge computations or
simulations.


> It may very well be the case that the 32 bit and 64 bit systems are
> entering a phase simular to the 16 bit -> 32 bit transition.

It sure is.  I understand the economic side of the argument, but the
technical side of this argument is, IMO, different than for the
16-vs-32.  My impression is that the shift to "managed" languages
plays important role, as garbage collectors have more bits to steal
from 64-bit addresses, so in the benchmark-driven industry dropping
32-bit support makes economic sense.


-uwe