sourcecode : binary :: ps : pdf

PostScript files are distilled (akin to compilation) into PDFs, but you can never positively reconstruct the PS when given only the PDF. PostScript is a Turing-complete programming language, whereas a PDF file is just the result of the compilation.

However, right now, I think your best chance is using a plain-text format:
* LaTeX. If you need to do an high-quality publication. There’s the problem of learning curve, though.
* An XML-based format like DocBook.
* A plain-text format like ReStructured Text or Markdown. If you don’t need to do too fancy things with the presentation, these will be extremely easy to learn for any collaborator and you can produce output in TeX, PDF and HTML (which can give great results with some good CSS).

I’m inclined to chose the last option for most stuff.

I generate and edit figures as svg files in Inkscape, then export (“compile”) them as pdfs for pdflatex. Committing the svg’s rather than the pdfs works well for tracking small changes (text changes within a figure, or moving objects relative to each other) but tends to look like blob comparisons for large changes. In this context (and probably the odf context as well) it would be useful to have a tool for putting an xml file in a “normalized” form to give, for example, the same branch ordering for similar files.

@greg re: legacy presentations. I think there are two useful assets here, both of which are not too hard to convert:
1) The structure and text of the presentation.
This can be converted to TeX or HTML by hand about as quickly as
you can read the presentation if you can touch type.
2) Reusable images and diagrams. Saving these as png and/or pdf
would be pretty straightforward from openoffice/inkscape and
would be sufficient to make them available for TeX/HTML
presentations.

I do both of the above to cannabalize my old odp presentations for my new latex-beamer presentations. These steps are not rate-limiting relative to thinking about the presentation and choosing the bits that I want to reuse.

I’ve switched presentation tools every couple years since 1999 (from marking up overhead transparencies with sharpies, to powerpoint, to HTML, to openoffice, to latex-beamer). The learning difficulties for each tool were pretty similar (ok, transparencies win), and generally easier than the graphic design, writing, and public speaking aspects.

@Daeng: I think that would work, too (or like Aslash suggests, TeX). But the main problem I foresee is converting Powerpoint/Impress files _into_ HTML/CSS (or TeX). Being able to use these other formats as a starting point is important because there is so much out there available right now under some Creative Commons license.

@kdt: I’ll ask about Chuck during our next update call. And of COURSE I will serve as Librations Copyright Specialist! :)

@Aslash: I’m not suggesting TeX because no one uses it in my line of work. And by “my line of work” I mean practically every line of work. Sad but true. Yes, it would solve the VCS problem, but solving that problem can not itself create a new problem of “everyone should switch to TeX to create presentations because I want to use a VCS.” See reasoning that I gave Daeng.

@Everyone: I guess I want to be able to take all of these legacy ppt/pptx/odp files and convert them into a HTML/CSS, TeX, or S5 type format, without formating loss, and with the ability to rebuild those ppt/pptx/odp files after modification. Is that so much to ask?!?!
(I think the answer is “yes”)

I love your description of binary blobs. I have been hoping you will serve as Librations’ copyright specialist, so I thought I’d mention that here, too. we’re going to need help choosing the right CC license… :)

The format, however, is very verbose and manual merges, even though now possible with a text or XML editor, may still be a pain.