DM communication standard

[Raffaele Sandrini]

On Mon, 2004-08-09 at 20:19, George wrote:
> FIFOs?  Evil! :)
> 
> > > George, Oswald: I think you should reconsider this issue. As a last
> > > resort, every app communicating with DMs could implement both, KDM's
> > > and GDM's way doing that, but that is not really a resonable solution.
> > > 
> > yeah, i expect gdm to adapt kdm's new way. :-]=]
> > i did this in a hurry without checking with george first, but judging by
> > the conversation we had before i expect george will have no real
> > objections against my stuff. the transport was already agreed upon and i
> > made the protocol extensible (see "caps" command), so gdm can add stuff
> > i left out in kdm (btw, "my" command set is not complete, yet).
> > oh, george, now that i was pointed at your existance again ;), would you
> > mind reviewing ctrl.c _r_sn, so i can pull the emergency brake before
> > our release if necessary? not that the stuff is already documented or
> > even used anywhere, but better if it's not there at all, if it's going
> > to change anyway - i left kdm's old fifos in for now.
> 
> I will take a look at it ...
> 
> But I'm really not keen on using fifo's rather then a single socket.
> 
> Secondly: this is just the DM, there are only two and it's really really
> really REALLY (did I meantion "really") low on my priority list to make gdm
> change it's protocol.
> 
> Since the only reasonable application that ever wants to call this is the
> session manager I don't see this as a big issue.
> 
> I think in the star trek future we should be using d-bus.  So this is the #1
> reason why gdm is very unlikely to adapt another incompatible interim
> protocol.  Basically, at some point when everyone and their mother is using
> d-bus, then I'm sure both me and oswald will go "oh yeah we should use that"
> and we'll implement that.  Since I think this will happen at some point, I
> consider any other change in protocol temporary and thus unnecessary and a
> complete waste of time unless it does something new and cool and I can't
> imagine it does :)
> 
> In the meantime I think gnome-session and the kde session manager should just
> implement both protocols.  That is far less coding then changing gdm.  And it
> doesn't break anything while changing gdm does.  If in the meantime either
> gnome-session nor kde session really implements the other, it is no biggie.
> All that happens is that the user will have to log out and then select "Shut
> down" from the login screen.  I think the 95% percent use case is when a
> gnome user uses gdm and a kde user uses kdm, so those are the cases that need
> attention now.  IMO anyway.

As there seems to bi little chance you two match in resonable time i
will most probabbly implement KDMs style seperatly.

cheers,
Raffaele