These are fascinating data files.  Version 1 is rather roughly
perforated and is resisting being cleaned up into a punch matrix,
but versions 2 and 3 are remarkably well cut and they dropped straight
into matrices.

It's immediately apparent that version 3 has been created by fiddling
with the master for version 2 -- although the musical effect is
distinctly different, there are little bits all over the place where
the rolls are exactly the same.  Putting the two side-by-side certainly
illustrates the roll-arranger's art very nicely.

There's a thoroughly-researched history of La Cumparsita at
http://www.todotango.com/english/biblioteca/cronicas/la_cumparsita.asp 
which describes how the piece was published in different forms over
the years, and you wonder if this is what inspired the differing roll
versions.

I've heard from several sources over the years that Pampa rolls are
rather good, and this little exercise confirms the point in an
intriguing manner.  The Pampa rolls (from Buenos Aires in Argentina)
seem distinctly better-arranged and better-made than the "Victor" rolls
from sort-of neighbouring Montevideo in Uruguay (where La Cumparsita
was written), a few of which I was recently loaned to scan.  I hope we
can organise a few recut projects -- the world needs more South
American piano rolls when they're as good as this!

"Muchas gracias" to Julian Zabaloy for making La Cumparsita available
using nothing but an A4 scanner.  That's dedication for you!

Julian Dyer