> So EEPROM is fast to erase, but then solving a software bug normally
> takes me more than the 8 minutes a /JW version takes to erase in UV,
> especially if you adopt a structured approach and try to test as
> much as possible before creating a new software build.

Comparing EPROMs with programmer-bound Flash/EEPROM devices you are right
that it's not much of an issue, though EEPROM devices can be sold in
inexpensive plastic packages unlike ERPOMs which must be sold in cerdips
[why not clear plastic?] and may be cheaper as a result [PIC 16C84 onezies
are about $6.50; 16C54 EEPROM onezies are more than twice that].

The PIC 16C84, however, has another substantial advantage over EPROM parts,
though: it can be easily programmed in-circuit.  When developing a system
where the CPU is not conveniently accessible, this can make a big difference.
On one project I did, swapping a CPU would probably take several minutes and
require shutting everything down [the CPU just handled a small part of the
overall system].  With the PIC 16C84, however, I was able to reprogram the
system to adjust its timing without having to power anything off [the system
was such that the PIC outputs tri-state to "safe" state].

> Finally I think that most large commercial users shy away from in
> service upgrades (although perhaps not in the military field), I'd
> prefer to think they get the bugs out before issue - when was the
> last time you found a bug in a video recorder !

Well, I guess that depends how you define "bug".  There are certainly enough
minor nits on my VCR that I wouldn't mind changing [e.g. there is no way,
sans remote, to select the composite video input; there is no way, other than
the "one-touch time/day recording" [which is on the console, not the remote]
to timer-record the composite video input [and that only after it's been
selected via the remote]; the channel-select buttons work in certain modes,
but not others; etc.] and if I could get a field-upgrade to fix those things,
I would.

TV set closed caption decoders, however, do sometimes have bugs [this is
especially the case in hotel TV sets.  Any idea why?]  The worst such bug
I saw was one which would process a doubled 142D code ("carriage return")
twice.  This rendered roll-captioning nearly impossible to read.  Further,
if television sets were field-upgradable, it would in many cases be a simple
matter to add new features [such as the "V-chip"]