Extra dashes and missing characters when converting unicode to ascii

I can reproduce the problem, but that table isn't in a useful form. Here's the code pathauto uses, borrowed from commentary in the online PHP manual:

  // Convert accented characters to their ASCII counterparts...
  $output = strtr(utf8_decode($output),
       "\xA1\xAA\xBA\xBF".
       "\xC0\xC1\xC2\xC3\xC5\xC7\xC8\xC9\xCA\xCB\xCC\xCD\xCE\xCF".
       "\xD0\xD1\xD2\xD3\xD4\xD5\xD8\xD9\xDA\xDB\xDD".
       "\xE0\xE1\xE2\xE3\xE5\xE7\xE8\xE9\xEA\xEB\xEC\xED\xEE\xEF".
       "\xF0\xF1\xF2\xF3\xF4\xF5\xF8\xF9\xFA\xFB\xFD\xFF",
       "!ao?AAAAACEEEEIIIIDNOOOOOUUUYaaaaaceeeeiiiidnooooouuuyy"); 
  // ...and ligatures too
  $output = utf8_encode(strtr($output, array("\xC4"=>"Ae", "\xC6"=>"AE", "\xD6"=>"Oe", 
    "\xDC"=>"Ue", "\xDE"=>"TH", "\xDF"=>"ss", "\xE4"=>"ae", "\xE6"=>"ae", 
    "\xF6"=>"oe", "\xFC"=>"ue", "\xFE"=>"th")));

Any thoughts on modifying this to take care of the reported problem cases?

Thanks.

I've edited pathauto.module a bit, to create translited aliases for russian node titles - but it doesn't uses these codes, it uses true cyrillic letters. So, I couldn't patch the original pathauto.module with unxutils.sourceforge.net diff.exe because of cyrillic letters, though I've created a patch.

and patch

This bug should not be seen as narrow as to fix only the accented characters. Many languages have a couple of special characters (ie. spanish:ñ, catalan:ç, turkish:ü,ö,ş etc.) that are also converted to extra dashes or just stripped. There must be a solution including all these.

One solution is to use utf-8 counterpart of the characters like Wikipedia is using. For example the url path of the term "españa" is http://es.wikipedia.org/wiki/Espa%C3%B1a

well, I think the easiest solution (the best?) is to increase the number of elements in array

 $output = utf8_encode(strtr($output, array("\xC4"=>"Ae", "\xC6"=>"AE", "\xD6"=>"Oe", 
    "\xDC"=>"Ue", "\xDE"=>"TH", "\xDF"=>"ss", "\xE4"=>"ae", "\xE6"=>"ae", 
    "\xF6"=>"oe", "\xFC"=>"ue", "\xFE"=>"th"))); 

you can find all symbol codes in charmap.exe (win32)
cyrillic, accented, ...etc

if I have time, I'll create the patch soon.

Partial fix - I've implemented a translation using the table you provided, which seems to improve the situation but doesn't cleanly translate all the characters from your original example (Ăă Îî Ââ Şş Ţţ Iñtërnâţiônàlizætiøns). Any advice on further refinement welcome...

I'll commit this to cvs later today.

Since mike committed most of this, I'm going to close it in favor of this issue which discusses new ideas for the transliteration process: http://drupal.org/node/61815