Mark G. Forbes:
 
> ... 'null' character ... gave the printhead carriage enough time to
> return to the left hand side of the platen before the next character
> arrived. That's why CR and LF are separate characters

   Errr, not exactly.  As per the following sentence following the above,
CR and LF are entirely different functions, THAT«s why they are separate.
Even in "the old days", use was made of this on the one hand to issue
successive LFs without CRs for vertical "white space" and on the other to
create overprint characters.  ASCII art on a VDU is a stripped-down
version of what you can do on a teletype (and stored on *paper* tape).
Also, who«s used APL on a teletype?

> If you didn't add the null, the next character would print while the
> carriage was still moving to the beginning of the line.

  The "nul" on the Baudot machine was the "LTRS" (letters) shift
character.  Some machines were however rigged to LTRS-on-CR (or was it
LF?).  Typists were therefore trained to CR, LF, LTRS and in THAT order.
Contrast this with VDUs where the CR is instantaneous but the LF slow (as
it requires shifting a whole screenful of characters and clearing the
bottom line).  These require(d) the sequence LF, CR, NUL(s), or at least
one more NUL if the CR came first.  While the LF was processed, the CR
would be held in the UART and successive NULs quite possibly lost!

  For this reason, the old O/Ss provided (and UNIX still does) functions
to set the number of NULs sent after newline(, or CR,) and in what order.
 
Harold Hallikainen:
> The model 15 did it all mechanically.  Quite amazing!

  Some of the most beautifully engineered and thoroughly reliable and
durable machines ever made.  Albeit with a MAJOR storage problem, I have
a small collection of readers, reperfs and the later model Siemens 100
and 200 (Baudot) machines, a DECwriter etc.  Can«t bear to part with
such glorious machines.
 
  Now, the *not quite* 1.5 stop bits matter:  The half-bit tends to
"dither" the timing so that if the receiver accidentally synchronises to
the wrong part of the character stream, it will be tend to right itself
in fewer characters.  Well, that«s as I understand it!

  Holding magnets: The Siemens is rather clever: It uses one magnet, but
six armatures "offered" in turn to the pole piece, one per bit (the sixth
relates to the start bit).  Also, the early Creed machines used bi-polar
(polarised) drive magnets for faster response.

  Relevance to PICs?  Well, I shall look up my old interrupt-driven
state-machine UART code (from the Mi-Co-Co) with a view to publication.

  Cheers,

  Paul B.