Re: How about xsl:fo to xml usingh xslt

A thing to consider is that by the time some editorial matter has made
it into XSL-FO state, it has by and large lost a lot of the
editorially pertinent semantics, in exchange for a rich set of layout
semantics (not the same thing at all). To do a roughly reverse trip,
you wouldn't for sure know if two kinds of block that contain similar
typographical content are in fact two instances of captions, or if one
is a caption and the other is a photo credit, or what.

The task is a bit like trying to reverse the process from old-school
PostScript back to the thing that made it (perhaps QuarkXpress). Who
is to know that the 18pt red Palatino line of copy came from the same
QuarkXpress stylesheet as the next instance of an 18pt red Palatino
line of copy? Or whether it was some other QuarkXpress stylesheet
indicating a completely different editorial purpose, that merely
happens to give an identical result on the page (either by accident or
design evolution)?

All the PostScript* tells you is that there's a line of copy in a
certain position on the page, of a certain typographical
specification. Elsewhere in the file it'll tell you there's a line of
copy in a certain other position on the page, of a certain
typographical specification. Nowhere in the PostScript is there any
kind of connection between the two at all. Even though the viewer can
see that they look like 'the same kind of thing' on the page.

The same kind of thing happens with XSL-FO - you mostly lose the
grouping that stylesheet classes at the editorial stage prior to
transformation might imply, in exchange for layout semantics.

*(ignoring PDF, particularly the current tagged and meta'd PDF 1.5
level, of course, which piggybacks a considerable amount of meta info
along with the PDF data, which would make it somewhat reversible back
to an editorial stage.)
-- 
Ian K Tindale
http://tindale.dyn.nu/