Re: Backspace, Delete and other keys

[Johnny Billquist <bqt%softjar.se@localhost>]

Greg A. Woods wrote:
> At Wed, 31 Mar 2010 22:07:05 +0200, Johnny Billquist <bqt%softjar.se@localhost> 
> wrote:
> Subject: Re: Backspace, Delete and other keys
> > Greg A. Woods wrote:
> > > See, for instance, <URL:http://en.wikipedia.org/wiki/Backspace>
> > Whoa! That page was rather revisionistic.
>
> :-)
>
> And worse they completely forgot to mention the use of backspace with
> whiteout, which being an English-speaking Canadian, was the primary
> reason I used the backspace key on a typewriter.  :-)
>
> I suspect this meaning of using "backspace" to correct a mistake is very
> widely held, at least in primarily English-speaking regions.

Probably.

> > Someone should send them the layout of a VT100 keyboard.
>
> No need -- I've got a real one in the garage. :-)

Not you. Them!

> > > Sadly some East-coast folks were no help when they decided that
> > > <Control-H> was distinct from <Backspace>.  :-)
> > Hmm. I don't know at all who you are referring to now. Please enlighten me.
>
> Unless I'm mixing my metaphors the wrong way I refer primarily to GNU
> Emacs and the desire to have the "help" function bound to the somewhat
> logical <Control-H> key sequence, but I think it was more a MIT thing
> overall, which also influenced Digital Equipment Corporation's
> engineers.

Ah! That came long before GNU-Emacs, though.
But I think you can blame RMS anyway, since he was the author of most of 
the original Emacs as well.

But in general, DEC *never* used BS to delete characters, and none of 
their terminals sent that code when you used the canonical key to delete 
to the left.

The VT100 indeed had a backspace key, but I don't think any other of 
their terminals even had that (hmm, maybe the LA120, but I'm not 
entirely sure).

> (Personally I find <Meta-?> to be a far more meaningful key sequence to
> use for finding "help" in Emacs, thus I don't have any problem with the
> fact the backspace key usually generates an ASCII BS.  :-))

Meta-anything is tricky, in that it either requires that you use the 8th 
bit as the meta modifier, or else run on something that can pass more 
information than a serial connection can.

> > Yes, in the way that there is no <Control-anything> in the ASCII table.
> > That's just a way to generate non-printable characters on most 
> > keyboards. So most keyboards allows you to generate a BS, by pressing 
> > ^H, the same way that you can press ^M as an alternative to pressing the 
> > return key (assuming your keyboard sends a CR when someone press the 
> > return key).
>
> Indeed, but again the big problem being for those people who want to use
> the <Control> key as a separate modifier key, and ideally (i.e. without
> ASCII in the middle) I'm sure they would have kept all <control-> key
> values separate from the ASCII "control" characters (i.e. the keys with
> names like "return" and "backspace" would generate the similarly named
> ASCII control characters, while the <Control-> modified keys would
> generate some entirely different values outside of ASCII, but then they
> also wanted a <Meta-> modifier too, and thus they ran out of bits on
> systems with 8-bit communications lines and had be satisfied with
> overlaying the <Control-> modifier on the ASCII control codes.

Yes, you could put it that way.

> > But apart from that, much confusion stems from the fact that people 
> > don't distinguish between key labels, the codes they send, and how the 
> > OS reacts when receiving them.
>
> Perhaps, though I contend most of it is a relatively recent invention.

I disagree. :-)
People press keys on keyboards. Keys have symbols on them. Keys then 
generate some code, which is sent to the OS, and the OS somehow reacts 
to this.
But people tend to ignore (or be unaware) of this fact. What is 
important here is to understand that we are speaking of three totally 
separate entities, which needs to agree, for the right thing to happen.

(I'm going to respond to Magnus Erikssons post next, and expand more on 
this.)

> Also some confusion can be eliminated by being careful to use the
> correct terminology in this context, i.e. that used by tty(4).  I'll try
> to do that here, but I'm bound to fail.  :-)

It's an area filled with booby traps. :-)

> > Looking historically, most of these have sent the DEL character, with 
> > the exception of the "Backspace" key, which usually sends BS.
> > And most any system accepted DEL as the code to delete one character to 
> > the left.
>
>
> The vast majority of early ASCII terminals I have used over the years
> generated an ASCII BS (010) character when the key generally found in
> the same place as the "backspace" key on a typewriter was pressed, and
> generally the Unix tty(4) "erase" character would ideally be set to 010
> to make full utility of this key.

Cool. You and I have very different experiences of terminals then.
I've used ADM3A, various Volker-Craig terminals, just about every DEC 
terminal, Televideos, Telerays, Facit terminals, ASR33, and god knows 
what else. None of them have sent BS.
A few of them could possibly have been set up to select what character 
to be sent, so maybe a few of them could be made to send BS if needed 
(not that I've ever needed).

> > Unix originally had a totally different mapping. It used @ to delete a 
> > line, and # to delete a character (unless I just accidentally swapped 
> > them now).
> > Eventually people in general got tired of that idea, and wanted to use 
> > those characters in their input. And also, since they were using 
> > terminals common of that day, they shifted to using DEL to delete 
> > characters.
>
> Well, that original mapping came from the use of an ASR-33 and similar
> printing terminals of course, where it was more convenient to use the
> lone "rub out" key to kill the current process (send it a SIGINTR), and
> since the original tty(4) driver didn't try to update the printout to
> show the effect of the erase character (over-striking would be kinda
> pointless anyway, if indeed the print carriage could even be backspaced
> by one character), a printed "#" was a clear indication to the user of
> how to figure out what was still in the input line buffer.  Note from
> V7's tty(4) that receipt of an ASCII DEL (0177) character was hard-coded
> and could not be changed, unlike sg_erase and sg_kill, which could be
> easily changed with stty(1) to be any character.

Yeah. Unix was always an oddball. :-)

> I.e. in Unix(tm) land, DEL has always meant send SIGINTR, and has never
> meant "erase" the character before the current position, and never could
> of course either, since its meaning was hard-coded in the tty(4) driver.

Unfortunately I don't have any history for stty, so I don't know how far 
back it reaches, but I thought these were all hardcoded in the early days.

Later on, they could all be changed. And I suspect it's a BSDism that 
you have "stty dec", which will set sane values for them.

> On the West Coast the Berkeley folks invented newtty(4) and there made
> the default "erase" character be ASCII BS (010), likely to match what
> common "smart" CRT-based terminals were generating from the key
> positioned near to where the backspace key on a typewriter was placed.
>
> So, no, Unix people did not switch to using DEL to erase characters.
> Most anyone who had any experience with early Unix was unlikely to ever
> switch to using DEL to delete previous characters -- they were very well
> accustomed to using it as the interrupt character!  It was only the
> people who started out with some experience with other systems where
> <Control-C> was the interrupt character, and who first encountered
> Unix-based systems _after_ the invention of newtty(4) or the somewhat
> similar tty(4) driver in Unix System III who sometimes chose to
> reconfigure their default settings to use ^C as their interrupt
> character, and only maybe then might they choose to use DEL as their
> erase character because some few terminals like the DEC VT100 _also_ had
> a DEL key conveniently located near to where the backspace key on a
> typewriter was located.

Well, you are probably right about the early Unix stuff. I didn't use 
Unix back then. I had totally forgotten about the mapping to INTR.
But, as noted, early terminals didn't even have a backspace key, but 
they did have a key that sent DEL.
So where this backspace comes from, I don't know.

> I.e. Don't blame any of this on the IBM PC for goodness sake!  It didn't
> help by mixing conventions so to speak (^C interrupts with ^H erase),
> but things were confused long before it came along.

Ok. I'll blame the IBM PC, and Unix then. :-)

        Johnny