IPv6 and localhost on F11-Alpha

Allen Kistler an037-ooai8 at yahoo.com
Fri Feb 13 00:16:22 UTC 2009


I discovered an annoyance on F9 that's worse on F11.  (I never checked 
on F10.)  I'd like to see what others think about changing some things 
in F11.

localhost should always be the loopback address, whether it's IPv4 or 
IPv6.  On F9 I can simply edit /etc/hosts (unowned by any package, 
created by anaconda) so that it is (for loopback entries):

127.0.0.1 localhost localhost.localdomain
::1       localhost localhost.localdomain

In other words, I could simply change the localhost6 names to get rid of 
the 6s.  Doing this edit is useful, at least, because tunneling X from a 
remote client over ssh (ssh -X) defines DISPLAY as localhost:10.0 (as it 
should).  But if sshd is running on IPv6, then the client app needs to 
be able to find a listener on [::1]:6010, which only works if ::1 is 
localhost.

On F9, editing /etc/hosts just works.  gethost* functions return both 
addresses for localhost, and apps figure out which one they need.

On F11-Alpha, the double entry for localhost also requires setting 
"multi" to "on" in /etc/host.conf (owned by setup.noarch), otherwise 
gethost* stops at the first localhost it finds.  (ping6 also seems to be 
an exception, but I'm not certain why.  I'm not convinced it matters, 
anyway.)

Note that there's an assertion in BZ#211800 that defining both 127.0.0.1 
and ::1 as localhost breaks some things, but doesn't say exactly what 
things.  The implication seems to be about programs that _change_ 
/etc/hosts (e.g., system-config-network) getting confused if there are 
both defined.

So...

1. What would be bad about getting F11 anaconda to create
    /etc/hosts with ::1 defined as localhost?
   (... except that maybe there's a bug in system-config-network that
   should be treated like a bug in system-config-network?  --  BTW, I
   haven't looked myself, yet.)

2. How come F11 resolution functions need "multi" defined when F9
    doesn't?  By that, I mean what's the vision shift in the compilation
    of glibc?  Is F9 the anomaly or is it F11?  Is the shift conscious?




More information about the fedora-test-list mailing list