Bug 616142

I have a request for consideration for a controversial feature that has two important and related use cases:

* I have a very large book to publish, but due to (various) reasons, only want to publish part of it. The content is all interlinked with xrefs, though, so I cannot build the book unless I manually change all of them.

* I am working on a single chapter of a very large book that takes up to 2 minutes or more to build (with publican 1/2+, which is still a huge improvement), and to view my work (to review as a draft, or to check the formatting), I would like to build a single chapter. However, due to the rigor inherited by xrefs, I cannot.

The best (and completely hypothetical) solution for these cases would be if xrefs were more like xi:includes:

For more information about firewalls, refer to <xref>
  <linkend>some_unique_id</linkend>
  <fallback>the <citetitle pubwork="section">Firewalls</citetitle> section of the Red Hat Enterprise Linux <citetitle>Security Guide</citetitle>.</fallback></xref>

I suppose technically that the mix of inline and block-level elements there wouldn't work, but that's the idea.

In lieu of such a useful construct, I am wondering if:

* there is any interest on behalf of others in building only part of a book;

* there are any suggestions for what to replace xrefs to nonexistent targets with; and,

* this is a good idea or a bad one.

This feature could be abused by users saying, "Well, if the book doesn't build because of broken xrefs, just turn on directive ignore_xrefs in publican.cfg." If this were a feature, publican could warn that it converted linkends A, B and C in X and Y chapters into ____ elements. Perhaps this could be a feature that could only be enabled on the command line, and thus could not set for entire books. I think its utility is undeniable, though, as it would save many of us time and effort. I currently cannot publish completed chapters of an overall 500+ page book I am editing (that is interlinked heavily with xrefs) without either changing all the xrefs manually in a publication branch, or writing a hack of a script that replaces these xrefs with <remark>s or some other element. I've done the latter out of necessity, but it is almost certainly better if this problem is discussed on the list, where we could perhaps come up with a single band-aid instead of many individual ones :-)