( Sorry for replying so late, the old Subject was not very appealing :-) On 23 Sep 1998, Manoj Srivastava wrote:
> I suggest that the preferred source format of the > documentation be also available. This means that we also ship > texinfo, tex, and sgml versions of the documentation, as well as HTML > formats (we may also ship man, info, ps formatted documentation too, > but the source and HTML should always be there). I would like to say that after modifying some of my GNU packages to ship the docs in HTML format, I dislike very much the idea of modifying the policy to ship the .texi source too. The problem I encountered was the following: I first planned to modify my debian/rules files so that a symlink for the .texi source is created from the debian/tmp/usr/doc/<package>-doc directory to the real file. Then it would be just a matter of cd debian/tmp/usr/doc/<package>-doc && texi2html -options foo.texi and removing the foo.texi symlink afterwards. Well, this didn't worked, because the .texi file included sometimes another files like "version.texi" and such. Therefore I had to create the html "in place" and move it to debian/tmp/usr/doc/<package>-doc afterwards. Some people propose that we ship .texi source in binary packages, but this means that we would have to identify *all* those extra files that are needed to generate the .html. This is a lot of (unneeded) work, IMHO. So shipping the .texi source will not be as easy as some people think, since there is not always a *single* file containing the "source". Sometimes there is even a Makefile to build the docs. Are we going to ship a Makefile too? It seems to me that the general rule that source belongs to the source package should apply here. We should already ship .html as this is our preferred documentation format (at least policy says so). Do we really need to make things more complex than they are now? Why don't we concentrate instead in finding a standard procedure for registering html docs, our (already) preferred documentation format? Thanks. -- "cccf08cbc9ee0529a5daf71e890f0901" (a truly random sig)
