Richard Robinson wrote:
| On Mon, 8 Jul 2002, Jack Campin wrote:
| > > Could we directly insert a little code at the beginning of an abc
| > > file to exclude it from your tune finder, like for the robot
| > > exclusion header in html files?

Actually that's not a bad idea.  I've been a bit busy lately; not
time to answer much email.  But I'll try to get arount to it.  But
there are the contra-indications ...

| > Mirroring is a major headache when using the Tune Finder - the way the
| > ABC corpus is now, you can get up to twenty copies of the same thing.
| > Something needs to be done to stop this getting any worse.  Google's
| > "do you want to see similar results?" is sensible.
|
| This is a problem, I agree. For John's TuneFinder, there would be the
| additional problem of identifying the primary version, rather than
| displaying an n-th generation copy and offering the primary as "similar"
| to it. This *could* be done, given universal agreement on a sensible use
| of headers. Umm, right ...

Yah; I'm not holding my breath until we get  universal  agreement  on
such  things.   There  is  one kludge that could help now.  We've had
occasional suggestions that we use the F:  header line to  contain  a
tune's  original URL.  People to mirror tunes usually wouldn't bother
changing this,  so  it  could  be  a  tipoff  that  the  original  is
elsewhere.   I responded to this suggestion by having the tune finder
generate a F:URL line showing where it got a tune from.  That way, if
you  ask  for an abc version of a single tune, it contains the source
URL, and you can use that information if you like.  I've seen  a  few
cases where tunes end up with several F:  lines as a result.

It did take a bit of experimenting to figure out how to  handle  line
terminators.   My  first  attempts  to generate F:  lines caused some
people's software to choke due to mixed line separators.   Finally  I
stumbled  across  what  seems  to  be the solution:  My code now just
canonicalizes on the ASCII standard (only LF as a terminator).   This
does  seem  to  work,  at  least  for people who download files via a
browser.  I'd guess that it's because browsers try to put text  files
into  their  system's  "standard"  form,  and  they  all  know how to
translate from standard ASCII to their system's form.

| A case in point ... I've just managed to put a big update to my "Tunebook"
| in place (less than ideal, because I'm going away tomorrow so won't be
| able to deal with the inevitable glitches and bugreports for a couple of
| weeks. Oh well, at least I've finally got the bulk of it done). Among the
| changes, it fixes a series of *horrible* typos to Skinner's "the Spey In
| Spate". These have been there since the thing started (?1994?), and will
| continue to be perpetuated by every site that's holding copies, unless
| they individually do something about it. I wonder how many of those people
| have thought of using mirroring software ? ... in my case, there are now
| headers that identify my version as mine, which TuneFinders could use, but
| they'd have to know about all the sites that do that, on an individual
| basis; and then there are copies that were taken before I put those
| headers in ... bleaargh.

One thing I've found impressive is how difficult it  can  be  to  get
typos corrected.  I sometimes get the feeling that people try hard to
locate the oldest transcription of a tune, and give that to all their
friends.   Of  course,  for  traditionalists,  this  isn't surprising
behavior.  But it  does  mean  that  an  unedited  version  tends  to
propagate faster than a later edited version.

Then there are the fun problems when files are  reformatted  slightly
as  they  pass from one kind of computer to another.  Especially when
you use email, and line wrapping and/or quoted-printable encoding  is
done at various steps.  Maybe some day this will be recognized as yet
another way that music can be transformed.

To subscribe/unsubscribe, point your browser to: http://www.tullochgorm.com/lists.html

Reply via email to