Joe,
On Fri, Nov 15, 2002 at 03:23:17PM -0600, Joe Wreschnig wrote:
> Don't CC me.
Okay.
> > Loss of information doesn't imply changing contents.
> If information was lost, you did, in fact, change the contents, from a
> legal and practical perspective.
In an ideal document, I have structural markup, so it is clear what is a
heading and what the author wanted to emphasize. With proper training,
you can read HTML like plain text and see the markup as just what it is
-- highlighting text passages. Unfortunately :-) most people cannot do
this, so they need some form of other presentation (=> style sheet). A
style sheet is considered good if it transforms the markup into
something the user can understand from her point of view and still sees
the original intent the author had (for us this means that headings are
properly structured, the text blocks are arranged that the reader follows
the original document's flow, etc.).
Now I postulate that I can change the style sheet as long as it fulfills
these basic requirements without "changing the content" of the document.
This means that I can make a text version from a PDF document by taking
the text in the order that I would read it if I read the PDF, mark all
bold sections with *stars* (note that the information that this was an
emphasis has already been lost), do proper section numbering so the
order of headings is clear, and so on...; then I can reflow the text to
75 columns, and I still have only changed the presentation. I may not
move commas, drop paragraph or (forced) line breaks, or reorder
paragraphs, of course.
> > and the authors
> > of the original papers were happy that they could point to me if someone
> > asked about a braille version.
> Good for those authors. However, that doesn't mean that the conversion
> was perfect.
If the conversion isn't perfect, this is a bug and should be fixed.
That's the only way we can deal with this.
The interesting point here is accessibility. Some documents may only be
available as .doc files and the like, in which case at least a PDF file
if not a text or HTML version should be provided so that I don't need
any non-free software to read the document. The document's license
should allow me to distribute such an HTML file, provided I don't change
the contents of the document. This can of course only be a best-effort
thing.
> I'm convinced that unless we can guarantee that all
> conversions are perfect, we can't formulate a legal means of ensuring
> that documentation is both freely convertable to and from any formats,
> and not modifiable except to change the format. You either have to place
> restrictions on the formats it can be converted into (thus making it
> non-free and probably obsolete about 5 years after publication), or let
> people modify it to change the format (in which case, there's no real
> reason you shouldn't make it truely free).
Yes there is. There are RFCs in HTML form out there. The wording hasn't
changed, obviously, or the file would be useless because I had to check
against the text version before using any information, and then I could
be reading the text version anyway.
> > When I'm talking about conversions, I don't necessarily mean automated
> > conversions. A lot of them will require at least some manual attention
> > (iso-8859-1 to ASCII has language-dependent rules for conversion, like
> > German "ä -> ae" or Danish "å -> aa").
> I'm familiar with these rules, too; are they actually formalized
> anywhere, or have they just arisen as conventions?
Good question, I don't really know; especially the case of the German
sharp s ('ß') is interesting: It is transcribed as "ss", but by
definition it is a ligature of "sz". I believe these are widespread
enough so we can use them safely, whether formalized or not.
> > The end user usually does not see whether this is a Quark or LaTeX
> > document, but instead, she gets some preformatted version (this is also
> > true for DFSG free documentation -- most people will prefer the info
> > browser or printing the PS to reading the texinfo source). So while such
> > a conversion would change the structure and everything, it would only
> > slightly change the end user's perception of the document -- if the
> > conversion is done right and the content unchanged.
> Until that end-user wanted to modify the document, in which case they'd
> go to the LaTeX or Quark or DocBook or whatever source.
If the license permits changing a document, then the "source" would
obviously be included. Bear in mind that these rules are basically
written for standards documents, which aren't usually changed.
> Let's see what
> happens when we apply your argument to programs:
[...]
Programs are different. For programs, the source is the legible form,
for documentation, it is the compiled form.
> Would you claim that, therefore, Debian should include proprietary
> software?
We should have a legible form of everything. This means source to
programs and documentation in text, HTML or PDF.
> > Note that this is only formulated that way to cater for standards
> > (again). Since all descriptive documents (which are the majority) are
> > supposed to be modifiable anyway, this clause won't apply for them, and
> > you just can convert them as you wish.
> IMO normative sections are the ones I'd want to modify/reuse the most,
> because they contain the information pertinent to the actual standard
> I'd want to extend or change.
So write a new document containing the changes. The guidelines give you
the right to reference and cite, and a list of changes is more useful to
people familiar with the original standard anyway. Besides, this is just
like requesting that modified versions have a different name, which is
acceptable (though not necessarily encouraged) by the DFSG,
> Why are you so intent on allowing blatently non-free documentation into
> Debian?
My primary concern are standards. There are few documents describing
standards that meet the DFSG, even a lot of the FSF's documentation
fails here (see the license of the gdb manual). It is unlikely that this
information gets relicensed, and I fail to see actual need for
relicensing as well. The documents are redistributable and
patent-unencumbered, thus usable. If John Doe were to modify them, I'd
first have to find out what is the actual standard and what not, thus
making them less useful. Besides, the only change from the DFSG is that
I dropped the requirement that changes be machine-readable.
Simon
--
GPG Fingerprint: 040E B5F7 84F1 4FBC CEAD ADC6 18A0 CC8D 5706 A4B4
Attachment:
pgpA11BGIsBfx.pgp
Description: PGP signature