In my current Texinfo manual I read:
For the printed output, you may specify an explicit sort key for an
index entry using '@sortas' immediately following the index command.
For example: '@findex @sortas{\} \ @r{(literal \ in @code{@@math})'
sorts the index entry this produces under backslash.
To reduce the quantity of sort keys you need to provide explicitly,
you may choose to ignore certain characters in index entries for the
purposes of sorting. The characters that you can currently choose to
ignore are '\', '-', '<' and '@', which are ignored by giving as an
argument to the '@set' command, respectively, 'txiindexbackslashignore',
'txiindexhyphenignore', 'txiindexlessthanignore' and
'txiindexatsignignore'. For example, specifying '@set
txiindexbackslashignore' causes the '\mathopsup' entry in the index for
this manual to be sorted as if it were 'mathopsup', so that it appears
among the other entries beginning 'M'.
Now obviously this might _finally_ make it possible for us to clean up
indexing and have entries like \tweak sorted under "t" while the index
entry properly is '\tweak'.
I don't know to what degree the HTML index would be affected. This
would require a reasonably recent version of Texinfo and/or texinfo.tex.
I believe that Gub is still stuck at version 4.something because of
efficiency considerations (Texinfo 4 is written in C rather than Perl
and thus faster). This indexing issue, however, might well be a
reasonable incentive to bite the bullet and move to a current Texinfo
version.
Thoughts?