> >On Thu, 18 Mar 2010 00:19:37 +0100 I. Szczesniak wrote: >> On Wed, Mar 17, 2010 at 2:57 PM, Glenn Fowler <gsf at research.att.com> >> wrote: >> > >> > On Wed, 17 Mar 2010 14:48:50 +0100 Casper.Dik at Sun.COM wrote: >> >> >what happens for (4)? >> > >> >> Failure of what? getexecattr()? >> > >> > I wasn't sure if there were other failure modes >> > this is covered by the NULL getexecattr() return > >> Careful. getexecattr() may do NIS/LDAP lookups and rivals (in a >> negative meaning) DNS lookups in performance. I don't think this is a >> usable way unless the values are cached. > >ksh upstream got boilerplate code way back to make ksh93 pfexec aware >we assume pfexec performance regressions would be the same for all shells > >if getexecattr() is really a hog then provide a cached version and we'll use it >
I don't think it is important; profile shells are not common shells. An additional binary is called at this time and it doesn't have any cached data other than available in nscd. Casper
