> I'm not quite sure of the advantage of breaking out a 75K texinfo file,
> but perhaps the author has something in mind for the next release.
This was one of those decisions which are not that easy, as either way
you do something wrong. The most correct way of dealing with this is
to just include the .texinfo file in the package. However, this
includes another dependency in the toolchain - everyone who builds the
package has to have makeinfo installed just for the docs. And this is
not something which could be regarded as optional, IMHO.
On the other hand, redundantly including any auto-generated file along
with its source (i.e. including the .texinfo and the .info) in the
package is something I try to avoid whenever possible, for aesthetical
and efficiency reasons.  (What I really hate is when people
either include the .dvi _and_ .ps _and_ .pdf versions along with .info
and .texinfo, or write Makefiles which force generating all those
versions when you just want to compile the program, and perhaps even
fail on the print formats when the proper tool is missing.)
 One exception I've already made in another project is gperf
output. This is a tool which is by all measures too exotic to require.