Re: exec and EXECSERVERS

[Thomas Bushnell, BSG]

Roland McGrath <roland@gnu.org> writes:

> > Why is that?  If it's programs that call setuid(getuid()) that have
> > this responsibility (as the original poster suggested), then this is
> > just fine.  On the other hand, my vote is that it's the setuid program
> > itself that always has the responsibility.  
> 
> That is a new responsibility that individual programs never had before, so
> programs not written with the Hurd in mind will not do it explicitly.  For
> the various variables that affect libc, the dynamic linker removes them
> from the environment.  So we could have it remove EXECSERVERS too.

Well, I think that the exec server should remove EXECSERVERS, on the
ground that it's the exec server that knows about the feature too.