Re: Qwerty, Retinas, & SGML

[John Walker <acad!throop!kelvin>]

While I agree with your observations, I can't really consider them
depressing.  Coming to terms with reality isn't depressing: it's liberating.
Highly-evolved systems almost always look like a mound of kludges when
viewed through the eyes of a competent designer--two years ago at the
closing session of Hackers' I tried to make the point that FORTRAN 77,
the Unix shell script language, the 3090 instruction set, etc., differ
from most academic designs in that they encode a large body of human
experience, certainly not optimally, but absent from them is the arrogance of
"users don't need that and we won't give it to them; we know best" which is
the central hubris of the "clean sheet orthogonal design".

Efficiently deploying capital and manpower in a large technological society
is a bitch.  Schemes which maintain a quasi-liquid market in corporate
control mediated by debt underwritten take-overs of corporations with minimal
ownership by their hired managers sound ridiculous also.  Anybody for
replacing it with an Office of Corporate Priorities and Central Production
Planning Board?

I've always admired the retina (by the way, the octopus retina design isn't
upside-down) because it's a classic example of ``cheap hardware, fancy
software'', an approach whose growth has tracked the development of cheap,
powerful computation.  RISC is all about ``Hell, let the compiler worry about
it--let's make this sucker cheap and FAST''.  The retina can be viewed as
``Hey--what's a brain for anyway?  We can edit out the blind spot in
software.''  Turning it photocell side out may result in poorer cooling,
higher probability of damage from transient high-intensity sources, and
less temporal resolution due to longer latency due to poor blood
supply (all of these are the purest speculation, but engineering tends to
work out in these kinds of directions).

Besides, if it weren't such a kludge the Argument From Design would be
more credible.