> And one thing computers are good at is trying the same thing again and
> again. Yes its minor but its very exploitable on a system with directory
> notifications or over NFS where transactions are slow.

Yeah, but so what? I have much more pressing things to worry about
than the small risk that somebody's private documents might be exposed
by a weakness in ispell.

> ispell is invoked on untrusted third party documents. Ispell is invoked on
> email being forwarded. Ispell is very worth auditing. Before you hit
> spellcheck when you reply to this email just think.. it might contain data
> designed to attack ispell. Its *that* easy....

In the first place, it can't -- I can see 100% of the characters. In
the second place, I can state with high confidence that ispell doesn't
have any buffer overflows with respect to the text being checked. I
got rid of those years ago, because they caused crashes and bugs. The
only place ispell is likely to have buffer overflows is in the
processing of command-line arguments or the hash file. For the
former, a wrapper script is sufficient (as I said). For the latter,
again I think that it's not a top-priority worry.

If somebody wants to send me a portable and stylistically appropriate
patch for either the mktemp weakness or any buffer overflows that
might be present, I'll happily apply them. But I'm far more concerned
with the lack of simple configuration than with security problems that
can't produce a root compromise.

-- 
    Geoff Kuenning   geoffcs.hmc.edu   http://www.cs.hmc.edu/~geoff/
One could not be a successful scientist without realizing that, in contrast to
the popular conception supported by newspapers and mothers of scientists, a
goodly number of scientists are not only narrow-minded and dull, but also just
stupid. -- James Watson

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]