Subj : Re: Hi all!
To : Digital Man
From : tenser
Date : Tue Sep 02 2025 01:19 am
On 31 Aug 2025 at 01:54p, Digital Man pondered and said...
DM> [snip] When I was building
DM> computers, it was still just a lot of assembly work, but you still had
DM> to know how to use 'debug.exe' to invoke the expansion ROM firmware of a
DM> "Winchester" controller, know how to low-level format a drive, know the
DM> differences between MFM and RLL encoding, platters, tracks, cyclinders
DM> and clusters and why it might matter for the customer, etc. Chips and
DM> cables weren't "keyed" and you had to know where pin-1 was and why it
DM> mattered.
It's kind of shocking to me sometimes how much of my accumulated
knowledge is utterly obsolete. Disk geometries are a good case
in point; practically nothing I use these days has spinning rust
attached to it anymore: I've got a large capacity backup drive
for Time Machine, and the Alpha down in the basement running VMS
has SCSI disks...but I really need to replace those. Anyway, it's
solid state devices all the way down nowadays, so the idea of
building software abstractions around rotational latency and so
on just feels absurd.
DM> Nowadays, all those details are abstracted away from the
DM> system builder.
Well, and the hardware, too: back when disk controllers on PCs
started to get smart enough to understand logical block addressing,
we stopped caring about cylinder/head/sector.
Much of this is good; consider PCI interrupt routing. Instead of
four level-triggered interrupt lines that require magic to discover
how they map to a physical line on a 8259A or IOAPIC, and because
they are separate signals from memory may outpace DMA, MSI/MSI-X
over the memory fabric is both much simpler and more rational. I
consider LBA similarly a strict improvement over CHS, and even
NVMe is much saner than SATA+AHCI. Memory-mapped IO accesses for
PCIe config space beat the pants off of the legacy port-based ECAM
stuff.
The biggest place where I think we've regressed is graphics. That
shit is weird.
DM> And the software stack is much higher now than back
DM> then, so the chances of one person knowing it all is even less likely,
DM> even when they do know enough to have a job in the field. I enjoy
DM> blowing the minds of youngsters when I'm able to demystify things and
DM> explain why things (in tech) are the way they are. But I also feel bad
DM> that they may not really retain the knowledge since they didn't "live
DM> it" and that could be a big handicap for the generation(s) taking over.
I think it's sad that they don't get to get their feet wet in the
same way, but I don't begrudge them not having to suffer through
the garbage that the PC world spat out in the 80s and into the 90s.
And I think stuff like Arduino and the lower-end embedded RP2040
stuff can help folks get into the guts on the electronics side.
