🚀 go-pugleaf

THE SITUATION:

Installing GNU finger 1.35 with the SUPPORT_FINGERRC option will include
code to execute a user's ~/.fingerrc file, if it exists and it executable.


THE PROBLEM:

Isn't this a *BIG* security hole?  Yes I know, I can simply undefine this
macro and avoid the whole problem.  Anyway, I tried a few simple things to
see if I can break this feature to gain unauthorized access and other
privileges.  I was (happily) unable to.  This is not to say that it is safe
to use.  Someone else might find a way to break in.

My real problem is that in all of the README and doc files, not even once
was it mentioned that this might be a security hole.  You *know* that will
be the first thing people will complain, so why didn't you say something
about that?  Why didn't you say how this feature works exactly (I have
looked at the src.)  IMHO you aught to dispell some fears that will arise by
mentioning how safe or unsafe is it to use.  The supplied documents bluntly
avoided that issue.

Having said that, I want to thank you for adding this feature.  It is better
than the fifo hack that was posted a while ago to the net.  This is one neat
feature I always wished for.


MY SETUP:

- GNU finger 1.35
- Sun-4's: everything from 4/25 to 4/630.
- Compiler used: gcc 2.2.2
- OS version SunOS 4.1.2
- network configuration: class B, subnets, ciscos, "the usual" stuff.  Not
  really relevant to this bug.  We nfs-automount many directories.
- Patches: none.

Erez Zadok.
Mobile Computing Laboratory.
Columbia University Department of Computer Science.
---
A wank's morning starts one machine   | Arpa:   ezk@cs.columbia.edu
clock cycle after midnight.           | Usenet: ...!rutgers!columbia!cs!ezk
 -HebrewWank #1 (US meaning, not UK!) | Bitnet: erzus@cuvmb.BITNET