@table without @item

[Karl Berry]

    @table without @item is warned against, although it allows to have
    an indentation that is consistent with indentation in other @table
    in most formats, and it is the only way to get that indentation.

When I look at the actual usage of this in the gdb manual (@defvar
Value.address, around line 23000 in the source, page 317 in my PDF
output), I don't think they are trying to achieve consistency among
@table's with and without @items, because there are no nearby
@address@hidden to be consistent with as far as I can see.  (I didn't
try to look at all the instances of this usage, though.)

Rather, it seems what they are trying to achieve is an
extremely-indented list, so that the list of variables/whatever is
clearly subordinate to the preceding text, like this:

--------------------------------------------------
  The following attributes are provided:
  
      Value.address                    [Variable]
      ...
  
      Value.is_optimized_out           [Variable]
      ...
--------------------------------------------------

@quotation and @table result in identical indentation in the HTML and
Info output, as far as I can see.  They get what they want because the
paragraph indentation happens to be less than that.  It is the TeX
output where it makes a difference, because the @table indentation is
larger than the @quotation indentation.  But even with @quotation, the
@defvar's are indented more than the paragraph indentation (in all
formats, since HTML doesn't have paragraph indentation).  Just not as
much as they'd like, evidently.

Such a large indentation also has the consequence of potentially
generating more pages, making books more expensive, less information per
page when browsing the PDF, etc.  Sigh.

I don't really care to argue this with them ad nauseum, but if you think
it's worth one more attempt to get them to use @quotation, given the
argument above, you or I could write them about it.  That seems like (by
far) the cleanest outcome, since in general, the whole concept of tables
without items makes no sense to me.  We could certainly provide them a
patch, if the boring search/replace work is the actual reason Eli is
arguing so vociferously for it.

Another simple approach we could take to placate them is to eliminate
the warning unless generating Docbook.  Or even to manufacture a fake
item for Docbook and avoid the warning there as well.

Another approach to placation would be to officially support @item-less
@tables, add it to the documentation, etc.  I would find it hard to
write any convincing justification for it, but it could be done.

However, both of those have the downside of other people not getting the
warning when they did not actually intend an @item-less @table -- a far
more common situation IMHO, so I'm not much in favor of them.

Another approach to placation would be to give them a customization
variable they can set to turn off the warning.  That at least would not
affect anyone else.

Or we can decline to placate them and let them live with the warning.

The one thing I don't want to add is add a special cmdline option or
@command to turn off that particular warning.  That would be
complicating the language to support something that, in my opinion,
doesn't make sense.

Wdyt?

I'm attaching my test file FYI.

karl

tableitem.tex
Description: Binary data