[PATCH] Resolve Service address in DNSSD::RemoteService

[Jakub Stachowski]

Dnia czwartek, 6 marca 2008, Sebastian Trüg napisał:
> On Thursday 06 March 2008 20:45:20 Jakub Stachowski wrote:
> > Dnia czwartek, 6 marca 2008, Sebastian Trüg napisał:
> > > Hi all,
> > >
> > > please find attached a patch for kdelibs/dnssd which adds handling of
> > > service addresses. ATM we only get the hostname which is completely
> > > useless in my network since I get stuff like "local.local".
> >
> > Why is it useless. There can be machine named 'local'  after all. But if
> > you then try to ping local.local. and there is lookup error then yes,
> > something is broken.
>
> I don't know enough about this topic. I just wanted to make it work for me
> and thought I shared. :)
> The name lookups failed here and since avahi provides the address... this
> seemed like the easy and logical solution.
>
> > > Address lookup was disabled via AVAHAI_LOOKUP_NO_ADDRESS. Why, I don't
> > > know.
> >
> > I always treated hostname resolving as system resolver's job - in this
> > case nss_mdnsd module. AFAIK  recent glibc versions have this built-in.
>
> Why exactly? If Avahi already provides it...
>
> And how would I go about that? I tried to do the lookup via QHostInfo which
> failed.
> If this is a config problem I vote to work around it since IMHO the whole
> point of something like Avahi is quick and simple connectivity. If I cannot
> use that from a default Linux installation, what good is it?

I don't have anything against it.

>
> > > Now the address is resolved and stored in ServiceBasePrivate,
> > > accessible via the new address() method. I know that it would probably
> > > make more sense to only store it in RemoteService but the design sees
> > > it only as a trivial extension to ServiceBase without any new members.
> > >
> > > Any objections?
> >
> > Two things:
> > 1) What about mdnsd? mdnsd-remoteservice.cpp is patched to use
> > ResolveEvent::m_address which does not exist. In fact mdnsd does not seem
> > to be capable of resolving service and its host name in one shot. So I
> > guess when resolve_callback gets called it has to first resolve hostname
> > and only then emit resolved(true)
>
> stupid me. I added that and was happy that it compiled. I never thought
> about the fact that on my avahi-based system the mdns code would never get
> compiled. :(

It's still needed for macosx and windows compatibility. Actually I wonder if 
anybody tested in on these platforms.

>
> > 2) Why use QString as type address? I think more specific type like
> > QHostAddress would be better.
>
> Correct. I just used QString because it is used by the DBus interface
> (naturally) and it was the fast solution.
> In the end the main goal was to get this discussion started. That is always
> simpler with some code to debate about. ;)
>

QHost address has constructor taking QString so additional work will be 
minimal anyway.

> Cheers,
> Sebastian