As an option (which may be in fact the default), this mode should also treat all runs of "variable" elements by only making the FIRST of them with a non-ignorable primary weight [.0201].
If there are several "variable" elements, the subsequent ones will get the primary weight [.0000] instead, but they will preserve all their 3 other weights. So in this mode with this option for runs of variable elements: " " (one spaces) would collate as [.0201.020A.0020.0002.0020] " " (two spaces) would collate as [.0201.020A.0020.0002.0020][.0000.020A.0020.0002.0020] This option for runs of variable elements is useful to collate together, at the primary level two equal words that are separated by different numbers of spaces. This new "Separating" mode will also be very useful for plain-text searches (working at level 1 only), as a MUCH better alternative to "Shifted" which finds too many occurences by ignoring all word separations and all spaces in arbitrary runs). So when searching for "km/h" it will still find "km h", or "km-h", or "km h" (two spaces in the middle), or "km / h" Philippe.
