> > The approach Jarl and I have settled on for the time being (it's not the best, > but it will work for the pilot), is to put the PID's of each Piper process in > the user's $HOME/piper_info/pids directory, then from UIL->DL->BL->PL (top to > bottom) send a "kill <PID>" (not "kill -9") to the next lower layer. This > will allow each layer to begin its own shutdown, and that should be that > (unless a layer can't die). > > The other benefit of storing the PID's is to determine what is running, or > what has crashed, should things be re-started. > > I was a bit confused about what Jarl wanted me to do in the UIL and DL, > because I knew that Brad had already developed a shutdown/crash recovery > mechanism. Brad explained his mechanism in a recent message and suggested > that Jarl use the same approach for the other layers (all layers). > > My question at this point is for Jarl: Do you understand what Brad is > proposing for the whole system (that which is implemented in the UIL and DL)? > What do you think? I understand and I agree. The meganism between the BL and PL already IS like Brad proposed, and in my reply on Brads email I asked him to point me out where I should work on in the Python code to add such a meganism to the DL->BL communications. Also I must add that this corbatized pinging is not enough, you cant tell if a layer is crashed or termintated correctly. > > Jarl also brought up a problem with sending network information (an XML > "document") from the UIL to the PL. He has been getting errors from the PL. > Jarl, can you send copies of those errors to this list so that Brad and > Jean-Marc can see them? > I could, but I would like to see JeanMarc and Brad give it a try themselfs, because I dont really know what is an error and what is not. bye, jarl