About the future format for LFS (reworded)

Matthew Burgess matthew at linuxfromscratch.org
Mon Sep 1 09:37:45 PDT 2008


On Mon, 1 Sep 2008 12:13:17 -0400, William Immendorf <computerperson1 at live.com> wrote:
 
> 1. Use Relax NG instead of DTDs. They offer more flexibilty than DTDs.

William, I don't have a problem with us moving to DocBook RNG, but need to be convinced that the effort it will take is worth it.  By and large, we do try to keep up with upstream versions of our toolchain, including the DocBook schema.  However, this is a much larger change than the previous incremental 4.x updates.

Last time I tried this out in my local working copy, there were significant barriers to entry.  The easiest of these to overcome is the changes to the entity and attribute names that you mention later on in your plan.  There is a conversion stylesheet provided by the DocBook authors that will largely automate this, though given the relatively small size of the LFS book, it's easy enough to do this using sed et al (BLFS may well be a whole other can of worms).

The stylesheets will take yet more work, and on top of that, and the biggest barrier I think we have, is the fact that xmllint and xsltproc couldn't handle RNG documents correctly.  I've reported the bugs previously to the upstream author's, but haven't tracked progress on these recently.

So, I think there is considerable effort involved in getting the the upstream bugs fixed, the stylesheets made compatible with the latest schema and the book sources in a state where they successfully validate.

After spending all this effort, just to get us to the same state as we're in currently (i.e. can validate and render the book sources), what, definitively do we gain?

Please don't just say "flexibility".  Flexibility to do what?  How is this useful to the LFS project?  Why can what you propose will be possible with DocBook-RNG not be possible with DocBook-XML and our current toolchain?

Note that I do think it would be nice to move to DocBook-XML-5.0.  This would keep us up to date with upstream packages, and is fairly unintrusive.  It's also a nice stepping-stone on the way to DocBook-RNG, should we decide that's the route we want to take.

Thanks,

Matt.




More information about the lfs-dev mailing list