On Thu, Mar 11, 2010 at 05:24:02PM -0800, Chris Quenelle wrote:
> Nicolas Williams wrote:
> > Is run-time version selection required?
> 
> Environment variables have an inherent kharmic
> burden which must be overcome.  They are intrinsically
> evil.  They are the "global variables" of the execution
> environment.

I share your profound dislike of environment variables.  Environment
variables get you in trouble in many cases.  For a twisted, but real
example: when IPC is involved and the participating processes don't have
the same lineage; breakage quickly ensues.

I agree that environment variables should not be used by verexec, or at
least that users should not rely on verexec's support for some
"override" via environment variables.

I thought, however, that verexec was intended to be driven by a
configuration file.  (You'll note that I didn't mention environment
variables until you did :)

I think I see your counter-argument: a hard-link (or even symlink) will
be resolved more quickly than verexec (or ld.so.1) can open a config
file, read through it, make a decision and exec() (or mmapobj()) the
result.  Did I predict well?

Nico
-- 
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to