Followup to Norm's write up on numbering.

["Rowland, Larry" <larry.rowland@hp.com>]

I have been remiss in getting this out.  I hope this comes out soon
enough for people to review it before the upcoming meeting.

This suggestion is a follow-up to the message  

[2] http://lists.oasis-open.org/archives/docbook/200606/msg00003.html

Proposal for Extending Numbering Schemes

 1) Add @startingnumber to ordered lists and procedures.  This is of
    value where a short ordered list or procedure is
    intended to provide an alternative to a section of an existing
    procedure or list.  

    Example:
  
    To consolidate a set of application to a single server, where all
    application have the same settings of kernel parameters, use the
    following procedure:

    Step 1

    Step 2

         substep a

         substep b
To consolidate a set of application to a single server, where all
    application have the same settings of kernel parameters, use
    the following procedure:

         substep c

    Step 3

    ...

    Step N


    In the case where the applications require different settings,
    replace steps 2 and 3 with the following:

    Step 2
         Substep a

         Substep b

         Substep c

         Substep d

    Step 3


 2) Add an optional @linenumberstep (or something similar) that allows
    the author to override the global setting for line number
    increment.  This is not to say that the stylesheet should not be
    as intelligent as possible about when to use the step increments
    based on the length of the example.  This is to accommodate those
    special cases where the stylesheet rules are producing
    inappropriate behavior for the specific case being dealt with.

    Example:

    Two programs are being discussed, one a short program that
    interacts with a longer program.  The short program is just over
    the length of listing that triggers numbering by steps in the
    style-sheets, but the author prefers the emphasis on details that
    the line numbering provides when every line is numbered.  For the
    fragments of the longer program, the decisions on when to use
    steps in line numbers are considered appropriate.  The author
    chooses to override the step size (setting @linenumberstep to 1)
    for the short driver program, while leaving the default values in
    place for the fragments of the larger program.

    This is, perhaps, even more important where the style-sheets do not
    make any attempt at providing differing behaviors based on the
    length of the code fragment.  The author would be likely to want
    to override the step increment on more of the program listings.