Peter Rabbitson wrote:
Hello,
I would like to duplicate a concern about the proposed fix, voiced over
at the debian BTS
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=493488#29, before the
final version ships.
=========================================================
This fix breaks the following setup:
1) Server A provides openvpn connectivity to clients
2) Servers X Y and Z are configured as VPN clients and provide some http
services both to the outside internet and to any VPN clients.
3) The http services are configured in a way that mandates password
authentication via an SSL channel, except when communicating with other
VPN clients.
4) Server A supplies 'push "route hostname.of.[X|Y|Z].server"', because
the servers in question are development machines, which can (and do)
change their IP addresses rather frequently.
With the current "fix" point 4 becomes impractical, and now besides
updates to the dns (which are automatic) I have to update the server
config every time something changes (which unfortunately is manual).
========================================================
Eventually the best way to deal with this is to test for
ip_addr_dotted_quad_safe and is_special_addr, and then attempt a dns
lookup on the string supplied for route. If anything comes back - use
the result as the routed IPs. Otherwise warn and carry on. This would
also fix this long-outstanding (not mine) wishlist:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=237251
I agree that this should be fixed, probably via a --route-fqdn-pull
option (as suggested) on the client to permit DNS lookups on option
parameters that previously allowed them.
James