Re: [Evolution-hackers] PIM server synchronization and Evolution online/offline state

[Matthew Barnes <mbarnes redhat com>]

On Tue, 2012-04-03 at 10:40 -0400, Matthew Barnes wrote:
> g_network_monitor_can_reach() takes a GSocketConnectable -- which is
> just an interface that's implemented by several concrete classes like
> GNetworkAddress (based on host name and port number) and GNetworkService
> (based on SRV records), so I assume service will be taken into account
> when possible in determining the boolean result that would feed into
> EBackend's "online" state.

Seems I'm assuming too much.

Dan Winship advises me that g_network_monitor_can_reach() merely checks
if there's a route to the host, but doesn't actually connect.

Would it make sense to split the "online" state into two flags, perhaps
"host-reachable" and "service-available"?  The latter would reflect the
result of actually trying to connect to the service to see if something
answers, obviously only attempted if "host-reachable" is set.

These could both be class methods in EBackend so backends can tailor
them as needed.  For example, an HTTP server may well respond alright,
but with a "501 Service Unavailable" which should interpreted as FALSE.

Would this distinction be useful to backends?

Matt