On Jul 15, 2016, at 11:30 AM, Chuck Guzis <cclist
at sydex.com> wrote:
On 07/15/2016 11:10 AM, Liam Proven wrote:
We don't appreciate how much faster modern
PCs are than the old
ones, because modern PC OSes are so appallingly slow and bloated.
Reminds me of a conversation that I had with Greg Mansfield back in the
mid-80s when he was working for Cray. I was grousing about the time
spent recompiling the BSD kernel on a VAX 11/750, even when streamlining
the process through partial recompilation (i.e. compiling only those
parts needing it). Greg was working with, IIRC, UniCOS at the time and
confided that on an X/MP he didn't bother with partial
recompilation--there was no practical time savings realizable.
Flash back to 1975 when recompiling the STAR OS kernel on a dedicated
STAR 1B took all night--assuming that the machine stayed up that long.
When I first started working on the IBM S/23, a complete build took a week
(yes, 7 days?if we were lucky). Debugging and fixing was mostly keeping
a notebook of patches to applied to the previous build. ?fixes? were first
developed by patch and then actual source changes were made. We usually
spent a day just patching the ?fixes? when a new build was released because
what we had to do in a patch vs the real change were often different.
Eventually someone wrote a cross build environment for the Series/1 and the
build went down to overnight (yea!).
You may ask ?It was IBM why didn?t you use the S/370 mainframes??. It was
accounting. We could ?buy? equipment (Series/1 and the like) and it was a
capitol expense. We were billed (at a ridiculous rate as I recall) for Mainframe
time out of the department expense budget. The expense budget was very
closely monitored. The capitol budget not so much.
Kids have it so much easier now. ;-)
P.S. A full build for the board I work on (OS and creating the boot image) for
work takes < 1 hour. The firmware I?m working on takes just 2-3 seconds to
build! This is on a PC with a 3.2GHz Skylake i7 with SSDs. ;-)
TTFN - Guy