Re: Thoughts on VDR evolution - my 2 cent

["Ralph Metzler <" <rjkm@netcologne.de>]

Margit.Fiegert@t-online.de writes:
 > I would prefer an architecture like this:
 > 
 > * A kind of logical device driver that is responsible managing the
 > resources:
 >   Which of the applications should get access to which of the resources
 > at what time.
 >   Applications: see above, 
 >   resources: Sattelite input channel, TV output channel, Videotext,...
 > of one or more cards.
 >   Of course the connections and priorities have to be taken into
 > account. Videotext can`t change the network while VDR is recording, ...

We are already working on a resource manager.
But it is implemented as a server, not as a logical device.
You can connect to it via Unix domain and TCP/IP sockets. 
On the domain sockets it offers file descriptor passing for higher
efficiency. 

File descriptor passing with Linux still seems to be a little
buggy. All the code we found for this (including code explicitely for
Linux) does not work reliably, sometimes depending on kernel version, 
SMP or not SMP, number of penguins in bathtub, ...
It works for us now but if somebody has more experience with
this under Linux please tell us.

Please also tell us your ideas regarding what kind of resources should be
provided and what kind of clients should get which priorities, how
access control via TCP/IP should be handled, etc. 

 
 > * A kind of Display Manager
 >   Input: Textual description of Menue structure
 >   Process: Interaction with user via LIRC or keyboard
 >   Output: Commands to the diffrent applications
 > 
 > * One program per application, connected only by access to the former
 > two modules.
 > 
 > I did some sort of generic Display Manager in my job some time ago. I`d
 > offer to help with this here too.

MHP already describes a lot of this (if you donīt have the specs get 
TS 101812 from www.etsi.org). But you can of course do something
different which does not have all the overhead.


Ralph