Re: Accidentals: Unwanted naturals

[Kieren MacMillan]

Hi David,

> Whether we like it or not, there are going to be intelligent people
> who misinterpret Lilypond's requirements in this case, and will
> perhaps not want to believe that such sophisticated software
> could be so naive/stupid/buggy/whatever as to
> not take into account its own \key declaration.

Here's where the problem lies, and — with all due respect — it's  
(still) not with Lilypond, it's with the user.

A key signature — whether hand-written (e.g., on manuscript paper) or  
computer engraved (e.g., by Lilypond) — is simply (and always will  
be) a *shorthand* for notating pitches in printed form with  
accidentals on every (necessary) pitchclass/note. For example, I  
could write out Schubert's Impromptu in G-flat major *completely  
without a key signature* (i.e., "in C major"): it would simply  
require a huge number of accidentals in order to communicate the  
correct pitches to the performer.

Many times when I'm composing, I'll start without a key signature and  
only assign one later (in order to make the performers' job easier),  
or vice versa. A perfect example is the fifth variation of my  
commissioned quartet that was premiered last month: I composed it  
(i.e., on my manuscript paper) with a D-major key signature, because  
that was the central tonality of the variation; however, I ended up  
engraving it (i.e., generated the score using Lilypond) without a key  
signature, because it was chromatic enough that I believed  
readability would be improved by "accidentalizing" every note,  
instead of requiring the players to compute the "current" pitch class  
(with all the previous accidentals in that measure) against a two- 
sharp key signature.

Not only does that example demonstrate why key signatures exist to  
begin with, it also shows how Lilypond's text entry (and concomitant  
transposition and re-keying mechanism) saved me a huge amount of  
work: if I had to "un-key-signaturize" the entire variation — because  
Lilypond's text entry didn't have the pitch alterations "built in" to  
the note names — it would have taken me a painfully long time.

> this is not some stupid misunderstanding, but a
> genuinely different perception of the process.

Fair enough... but I posit that it is not Lilypond's place to teach  
someone music theory or how to think in the (standard) Western  
notation system, especially when it flies in the face of nearly a  
thousand years of notational practice.

That being said, nobody's stopping anyone who wants to write a Scheme  
function that would support such a beast: that's the beauty of open  
source software!  ;)
I would just hope (and certainly voice my desire) that the powers  
that be — Han-Wen, Graham, etc. — would not add such a function to  
the base distribution, nor give it any serious face-time in the  
documentation.

Best regards,
Kieren.