In contrast with Karl, I am not that afraid of other parsers, as long as
they check their results
In theory, that would be fine. In practice, everyone else who has ever
attempted to write a parser fails to maintain it over time --- let alone
do anything as rigorous as the checking you are talking about. Leading
to language skew for anyone who uses it. Hence my practical conclusion
that it is a bad idea to undertake.
And *within* GNU, as I had said, there is no reason at all to implement
a full parser, because rms's whole idea for GNU is one program to do one
job. (Docstring subsets aside, that's clearly different.) Replace
makeinfo, yes ... as we just did. A second Texinfo parser, no.
Not that I have any illusions that these words are worth anything.
People will do whatever they want to do, regardless.
k
