00:22:56 Thomas Juerges (LOFAR): I can barely hear Emmanuel 00:23:19 Zbigniew Reszela: I have the same problem 00:23:34 Graziano Scalamera: +1 00:23:43 Matteo Di Carlo: same 00:23:47 Andy Gotz: When I increase the volume it helps ... 00:24:22 Thomas Juerges (LOFAR): @andy indeed. but then all the other sounds are deafening loud. 00:24:44 Graziano Scalamera: yes 00:24:45 Thomas Juerges (LOFAR): much better 00:34:36 Andy Gotz: We propose to gather questions here (in the chat) which we can ask Emmanuel after his presentation. Put "QUESTION" in front of your question so we know this is a question and not just a comment. 00:38:14 Thomas Juerges (LOFAR): QUESTION Does receiving of an event in a callback function block the receiving of the next event? 00:40:52 thomas.braun@byte-physics.de: QUESTION Is anyone using here tango on big-endian architectures? 00:42:04 Sergio Rubio Manrique: QUESTION for all: anyone using here the Pull model for events? (client-side buffering) 00:46:25 Lorenzo Pivetta: @thomas.brown yes, at Elettra we're using PowerPC 00:49:28 Anton Joubert: QUESTION: Demo client - what was the environment variable we need to set to see event details printed on the console? 00:50:00 Andy Gotz: QUESTION: where is the time in the HEARTBEAT event? I did not see it in the tracing output. 00:50:16 bolkhov: QUESTION (to page 11): if ZMQ event distribution re-uses CORBA code for data marshalling, then how do you plan to proceed after prospective removal of CORBA from TANGO? Wouldn't this code reuse make de-CORBA'ing impossible? Or is a byte-to-byte-compatible reimplementation of CORBA marshalling planned? 00:52:11 thomas.braun@byte-physics.de: @Anton Joubert. It is probably ORBtraceLevel, see 4.1.3 in http://omniorb.sourceforge.net/omni42/omniORB/omniORB004.html (but I'm just guessing) 00:52:45 Matteo Di Carlo: I thought it was the logs of the admin device.. hmm 00:53:34 Reynald Bourtembourg: You're right Anton. And if I remember correctly, if you set ORBtraceLevel to 30 (Be careful lower case 't'), you should be able to see the events traces 00:53:41 Andy Gotz: @boikhov this depends on if we want binary compatible on the wire or not. If we do then we would need to only keep the CDR marshalling routines and not the whole library. If we replace CORBA then we will replace the coding on the wire and then we would replace CDR for the events too. 01:00:51 Sergio Rubio Manrique: is it possible to fix the port addresses? 01:01:16 Sergio Rubio Manrique: Sorry, QUESTION: regarding the PUB ports in the device process. is it possible to fix the port addresses? 01:10:14 Sergio Rubio Manrique: QUESTION: when pushing events from code, should we use a mutex/lock to avoid concurrent push() calls? 01:13:29 Sergio Rubio Manrique: ok, just answered on this slide 01:24:28 thomas.braun@byte-physics.de: QUESTION How can the event name be different on the client and DS? Does "event name" here mean fully qualified name? If yes, I think I understand as it includes the network address/name. 01:25:53 Reynald Bourtembourg: @thomas.braun: you understood well. event name includes the fully qualified name so it includes the network address/name 01:30:16 Graziano Scalamera: QUESTION: is it safe to subscribe events from the same DS pushing them (so a DS client of its own attributes). I am already using and it seems to work, Just some strange log messages are printed 01:37:43 Sergio Rubio Manrique: QUESTION: the ZMQ buffer would limit the number of events receivable per-device or per-client? 01:40:10 Sonja Vrcic: Question: Can you comment on application where a single client monitors a large number of clients ? Are there any known issues (and recommendations)? 01:45:27 Sergio Rubio Manrique: Hi Sonja, apart of ZMQ buffer size, subscribing to thousands of attributes at client startup has been an issue for us. Using the pull model as described by Manu at the beginning could solve that but we haven't tried yet 01:47:58 Sonja Vrcic: QUESTION: when a client is monitoring a number of attributes (events) for the same DS, is there a single heartbeat. Per DS? 01:50:36 Anton Joubert: QUESTION: Was wildcard subscription considered in the design? E.g., subscribe to change events on all attributes. 01:51:41 Andy Gotz: QUESTION to EVERYBODY: is the performance of events sufficient for your use cases? 01:52:02 Lorenzo Pivetta: @Sonia Vrcic: the alarm system tango device (C++) in our setup is usually subscribing to several hundreds (sometimes thousands) events from many devices with no particular issues, except a device startup time non so fast 01:53:25 Andy Gotz: @Sonia we could make it faster by removing the synchronous call 01:54:05 thomas.braun@byte-physics.de: Hmm the synchronous call does serve a purpose. 01:54:05 Jairo Moldes Fuentes: I have used it 01:54:52 Jairo Moldes Fuentes: I have used the pull model, I mean. 01:55:05 Andy Gotz: I agree not to remove it but we could add a second call which allows you to subscribe without the synchronous call. Up to the client to choose which one to use. 01:56:59 Sonja Vrcic: Thank you for the excellent presentation ! 01:57:13 ELATTAOUI: Thank you Manu! 01:57:15 Sergio Rubio Manrique: I agree with you Andy, may be useful 01:57:19 Sergio Rubio Manrique: Thanks for the presentation!