new layout for the book (Gerard?)

Matthew Burgess matthew at
Sun Jan 18 15:17:48 PST 2004

On Sun, 18 Jan 2004 22:35:44 +0100
Alex Groenewoud <alex at> wrote:

> Matthew Burgess wrote:
> > Hmm I've just downloaded it now.  To be honest, the very first thing
> > I noticed was it's not XML at all (well the XML that's in there is
> > surrounded by non-standard constructs) and requires python to render
> > it.
> The important thing is that everything about a package is in a single
> file: if you want to bump the version of sed, remove a patch, and
> tweak the instructions, there's only one file to edit, instead of two
> or three different ones.

Yes, I see that now.  One thing to watch out for though is that
installation instructions sometimes differ between chapters 5 and 6
(e.g. patches specific to /tools, etc.).  XInclude processing allows one
to reference a part of a file, rather than the whole of it, so you're
booss structure can be adopted if it's felt that it fits our needs.

> What you have done in the newxml is a reproduction of the current
> book,including most of its rigid internal structure, you only use a
> different tool to render it.  I don't know what to say.

Well, prior to your recent commits, whereby appendixa simply became
links then it made sense for the common info to be in there and then
pulled in from wherever it was required (e.g. in chapters 5 and 6). 
Therefore the dir layout seemed to fit more-or-less what was required
(IMO of course - I didn't have any other opinions to base my decision
on at the time!).

Now, on the assumption that we can get XInclude processing working
on just parts of a file, where do we put the full contents of each
package? I would suggest in chapter 6 simply so that *all* packages have
their full contents there, i.e. it's consistent. Chapter 5 then pulls in
whatever content they can from chapter 6, and if the instructions differ
these are simply inserted in chapter 5's package file instead of the
relevant XInclude instruction. Appendices A and B can be automatically
generated I believe(they're just links) so don't require any source
files at all.  Likewise, I think chapter 4 can be produced automatically
as the download links are already in chapter 6.

As far as the directory names are concerned, I still maintain that the
current structure is the best we can do, simply because if I notice a
problem in chapter 6 of the online book, the obvious place for me to
look in the sources is in chapter 6.  No mental mapping required.

Comments on the above are greatly appreciated.  I currently have very
little free time, so if anyone wants to play about with any of these
ideas then please do so and share your findings



More information about the lfs-dev mailing list