marcello said:

If we want to eliminate the widows and clubs we must   - include `stretchability' in the leading or   - have a ragged bottom.

paper-books use the first solution, almost exclusively. so that's the one you want.

Both solutions can be worse eyesores than the original problem.

untrue. that's why typographers use the variable-leading method.

With stretchable leading the lines on the left hand page will not match the lines on the right hand page.

and that's totally meaningless, which is they use that method.

With ragged bottom, facing pages may be of different length.

that problem is not _quite_ so meaningless, because bottom-balancing the page-spread is esthetically pleasing. but if the only way you can get rid of the widows and orphans is to give up bottom-balancing, then do it. the widow/orphan problem is _much_ worse, especially with some _pages_ having one line! nine out of ten typographers will tell you so. (and the tenth one is the worst of those ten.)

In commercial typesetting these problems are overcome by manually tightening or loosening some paragraphs on the page, or even making the author rewrite some of the copy to fit the page.

since the introduction of computers into typesetting, the manual adjustment of leading is now antiquated...

In a purely automated process this is impossible.

in a purely automated process, copy-fitting is automated.

We also want to keep the flexibility of changing fonts, page sizes etc. so we cannot insert manual fixes.

well, my program controls for widows and orphans, as well as bottom-balancing page-spreads, and yes, it allows the person to change fonts, pagesize, etc. so you won't get very far with that "impossible" line. but hey, if you and your tools can't do it, i understand. still, give up the bottom-balancing if you must do so, because widows and orphans will make typographers laugh at your output... -bowerbird