Dischi, The only reservation i have is leaving a stable webserver. We use a few features of twisted which are nice. Rewrite all images and styles references to the first level so we don't have to have image and styles dirs at every level, the cgi path way of sending variables in one special case with library to help with m$ browsers only believing file extensions rather than MIME headers, HTTP basic auth, and then they have a very nice mechanism for shoveling out the file data when things get downloaded. Hopefully if we can get these type of features or have the ability to hack those in it would be great. The N-Way communication of MBus is certainly attractive for the types of things we won't to do in the future. just my .02.
Mike Ruelle [EMAIL PROTECTED] On Mon, 2004-09-06 at 13:19, Dirk Meyer wrote: > Hi, > > some already know because I told it on IRC, but here the official > "request for discussion": > > I don't like twisted. Why? It's a huge monster, including not only > stuff we need like xmlrpc and a web server but also stuff we don't > need and a main loop that is too complex. I saw in the code that it > waits for dead children not started by any part of twisted .. Hey, > that are my children, don't touch them! The next problem is, that we > have one xmlrpc server in the recordserver. But when the schedule > changes, Freevo has to ask the recordserver for a new schedule, the > recordserver can't send it to Freevo because a) freevo has no xmlrpc > port and b) if you have more than one freevo, how should the server > know that? And we need a way to send stuff to Freevo, like the > webserver may want to say 'start this audio playlist please'. > > A solution to all this (and more): we use pymbus. Mbus is a protcol > for host or link local communication. As the name says, it's a bus. So > every programm on the bus can get the messages (you can also send a > message directly). For detailed information, please look at > http://www.mbus.org. A brief summary: > > o mbus is secure, it uses encryption on the bus. > o mbus is plain ascii (after decrypted), so every app with the key can > read it very easy and apps like mspy can be used for debugging and > apps like mbussh can be used to send commands to an app > o you get notice when an app joins / leaves the bus > o you can define names for apps so you understand what they do > o you have events _and_ client server rpc communiaction. > o Bindings for C/C++/Perl/Python > > Example: > Freevo starts, joins the bus and ask: > Freevo: I need recording informations. Anyone here can provide them? > > Case a: recordserver is running: > RS: sure, I'm a recordserver, I can record dvb and analog. Here is my > current schedule. > Case b: no recordserver, Freevo still has no schedule and will never > aks again. When the recordserver starts: > RS: Hi, I'm a recordserver, here is my schedule. > > Freevo knows there is a recordserver with the name RS on the bus and > the user wants to schedule something. > Freevo->RS (unicast): RS, please record xy > RS no everyone: my schedule changes: here is a new schedule > RS->Freevo: done! > > Next is the webserver knowing that Freevo is running: > WS->Freevo: please play playlist xy > > You have an isdn watcher and a new call comes in. It doesn't know > about Freevo at all: > ISDN: in case someone wants to know this: the telephone is rinning. > A special Freevo plugin notice that and mutes the background audio > playback. > > And we can also remove the bad hacked network interface, you can send > events directly to Freevo: > app->Freevo: Button 'DISPLAY' is pressed > or > app->Freevo: Event(OSD_MESSAGE, arg='hi') > > That's what mbus can do, and much more. Read the rfc on mbus if you > want to know more. The second lib pynotifier is a simple (!) main > loop. The author of pynotifier and the current maintainer of pymbus > (also using Freevo) also has a simple webserver and non blocking http > client. > > > So, there is no need for twisted, mbus is the future[1]. And the > author of pynotifier will include pynotifier in Freevo when he has a > free weekend. > > Comments? Someone with a good argument against it? > > > > Dischi > > > Footnotes: > [1] I better say it before you find out yourself: mbus was developed > at the UCL in London and the University of Bremen. I also work at > the University in Bremen and mbus was developed in our working > group. ------------------------------------------------------- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click _______________________________________________ Freevo-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/freevo-devel
