Le jeudi 20 juillet 2006 à 09:54 +0200, Mathias Bauer a écrit : 
> Nicolas Mailhot wrote:
> 
> > Is there a way to take an OO.o doc, get all letters replaced with X or x,
> > metadata stripped, embedded images replaced by blanks with the same sizes,
> > and every other names (variables, bookmarks, fields, references,
> > color/style names) anonymized?
> 
> Replacing the text should be easy, also stripping the metadata.
> Replacing the images might be a bit harder, perhaps creating an empty
> image or metafile of the same size and replacing the embedded one is
> possible through direct file access. Another approach is replacing
> embedded images by links (to anywhere). 

You need to replace images by blank images of the same size or you'll
hide pagination bugs linked to image size (not a theorical concern - one
of those things is why I initially wrote this)

Also a 2-color black or white png compresses well, so that would help
documents fall under the upload size limit.

> I assume that the document
> should behave as before even with the broken links.

I fear in many times the pagination would change

> The problem is: what else needs to be exchanged, what is really
> necessary and what's just paranoia?
> 
> So we need a complete list and we need to dicuss what needs to be on it.
> As an example, why do you mean that color and style names need to be
> replaced?

because humans will choose descriptive names which may include company
name, and paranoïd users won't upload documents which may link them to a
particular employer (see the Sun foo colors in the default palette)

> > As most bugs happen on complex documents, most complex documents are
> > created in corp-space and corporations don't like disseminating internal
> > info for debugging purposes I suppose I'm far from the only one with
> > knowledge of bugs but no way to report them.
> 
> Yes, I totally agree. We could get much more (and better) bug documents
> and of course that would be fine.

Regards,

-- 
Nicolas Mailhot

Attachment: signature.asc
Description: Ceci est une partie de message numériquement signée

Reply via email to