This is the mail archive of the libc-help@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Tracing Linux hostname resolution


  Hi!

On Tue, Jan 25, 2011 at 07:02:35PM -0500, Jeff Blaine wrote:
> >   The 'hosts' service maps to the "obsolete" interface. Use 'ahosts' .
> >But frankly, glibc-2.5 really is ancient (as in, about five years old?),
> >any reasonably modern distro should carry something much newer.
> 
> Well, it's what ships with Red Hat (and CentOS) v5.5, which is
> *incredibly common* in an awful lot of production environments
> (not desktops...).

  Sure, but pointing out $whatever about such glibc (outside of RH
bugzilla) won't change anything since nothing is going to change about
it either.

> It's no matter, though, as Andreas pointed out that 'getent' is
> a tool only for debugging purposes, apparently.  And I pointed
> out that it should be indicated as such in the man page if that
> is the case.
> 
> strace-ing a simple ping's host resolution made everything clear
> again (instead of getent).

  I don't see where Andreas pointed out that it is only for debugging
purposes. At any rate, I agree that it is not a very advertised
interface; I'm not sure there is any official statement about its
support level, though.

> >   Note that it's not clear why are you so worried about host.conf; only
> >the 'order' keyword is obsolete. In particular "multi on" is fairly
> >important to get expected /etc/hosts behavior, esp. in multi-AF scenario.
> 
> I stand corrected.  Personally, I never understood the point of it,
> but that is only because in common use throughout the years, I
> have never seen anyone use anything other than "order" (which is
> now redundant when using NSS).  I did search around for more
> information and that left me finding only more people who really
> have no idea why it exists anymore.  It is clear, finally (which
> should have been the first step) viewing the man page, that it
> serves a role as you have pointed out.

  Well, the servers *I* have access to (SLE, Debian - together, also
should qualify as "incredibly common") almost all have 'multi on'. :-)
Thus, I'd say RHEL is more of an exception.

-- 
				Petr "Pasky" Baudis
Computer science education cannot make an expert programmer any more
than studying brushes and pigment can make an expert painter. --esr


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]