Re: "no software company or its customers could reasonably afford these days"
"I think Microsoft could afford a proper manual."
Back in the day, they did. I still have a copy of the Microsoft Windows NT System Guide vintage 1993; a task-oriented guide to how to do things in Windows NT (3.1? It's not obvious). An actual user manual. I also had a complete set of manuals for Office Pro (?) from a similar era - proper printed manuals for Access, Excel, Powerpoint, and Word. Junked them recently, kept the System Guide as a reminder that the business end of MS used to understand what was needed.
VMS manuals have deservedly been mentioned, but actually the DEC hardware and software manuals in general were in a class rarely seen in recent years. The focus was on content and not presentation, though once they moved from Runoff to VAX Document (via some others) even the presentation was good. Authoring a quality document was apparently easy enough too, unlike with (say) Word.
Manuals written by people who knew their product and knew their audiences (end users, system managers, programmers, field service engineers, even potential purchasers - different audiences with different needs) and knew how to write comprehensible english. Even the presales literature was worth a look. What a concept.
Is proper vendor-produced documentation inconceivable today? I've been impressed with what I've seen from Xilinx when I briefly played with FPGAs - are they really OK, are there other examples of how software/system documentation should be done, is it just MS that see it as yet another revenue opportunity for themselves or a partner?
The MicroVMS documentation set deserves particular mention. MicroVMS was just a repackaging of VMS targetted for MicroVAXes, around the VMS V4 era, and it came with a little known and short lived custom docset which disappeared once VMS on CD arrived and DECwindows and CD documentation came with it. It was a very small number (basically 2 manuals plus administrivia such as release notes) of compact manuals focused on the things that were most likely to be useful to most people (so not a complete reference, but a very usefuly desktop library, for small desktops).
 MicroVMS User's Primer AA-Z210C-TE 1985 partly available at
MicroVMS Users Manual Parts 1 and 2 (not online? boo hiss?)
El Reg: This eight day old article is still open for comments. Earlier today I wanted to comment on a two day old article  and comments are already closed. What gives?
Fwiw, what I wanted to say there was that USB->serial adapters are only OK if your requirements are simple. If your requirements are more complex, e.g. require low latency/rapid turnround, generic USB-serial may bring challenges, e.g. simple ping-pong protocols suddenly take N (e.g. ~5) times as long to compete a given task, because the latency between port and app goes from zero to multiple milliseconds, and if your protocol is immutably set in stone by (e.g.) a machine vendor and there are literally millions of pings and pongs because what you're doing is downloading a multimegabyte program byte by byte before you can process the next item on the production line... well USB to serial isn't helpful.
Relevant here too, because you will largely find this out by bitter experience rather than by reading the manuals. FTDIchip's USB->232 docs may mention this, but how many folks are aware of FTDIchip let alone the USB latency/turnround issue?