Robotics Service Bus

h1. Meetings2011-06-30

{{>toc}}

h2. Use-case Analysis Presentations

* Stefan
** Spread cross-compiling does not work well
** RSC more fine-grained because one single library to load wastes memory
*** build all smaller libs with one CMake but configure via options what is required
** Make tests and spread binary finding optional in RSB
** How to map CAN messages to RSB messages (Shadow Hand uses a more complicated protocol than Stefan's robot)
** Performance requirements:
*** 1 Mbit on the CAN bus
*** Marshaling could waste CPU
** Robert and Stefan will further discuss what is required to build a more or less generic CAN to RSB bridge
* Florian
** Sending and receiving in a gstreamer pipeline using RSB
** Audio missing
** Overview about available classes
** nearly easier than XCF
* Christoph
** high-level data, IPC
*** how to use or define common types for this (type-safety vs. flexibility)
**** Frederic: e.g. starting from BTL and ROS messages
**** make a publicly available repository that is visible
*** migration path from XCF
** video
*** for one computer there's a shared memory solution existing (later to be ported)
*** common types for images as protocol
**** support compression (e.g. run-length coding)
*** avoid extensions to this type(?)
* Robert
** using pbuf code not in whole app, how to convert to real domain types?
*** 2-step procedure
**** protobuf - network, centrally shared to provide compatibility
**** protobuf - domain, several ways because several domain objects with different dependencies, really centrally shared? - dependency hell?
*** fast decision required
* Johannes/Jan
** Conditional Sending (see above)
*** Problem statement and requirements
*** Implementation Plan
** Inter-Segment Communication
*** Clarification
*** Documentation: wiki page

h2. Additional Topics

* Client/Server Communication
* Inprocess and Shmem Communication
* SynchronizedQueue:tryPop() -> bug report
* XCF Migration
** Translation between XML and protobuf?
*** XML less desired
*** automatic translation based on XML schema could be a solution
*** protobuf probably suitable to represent all existing concepts like XML
** XPath like is a really desired feature
** CTU compatibility layer
** Active Memory

h2. Outlook: Next Meeting(s)

* Thies
* Hendrik / Ramin
* Introduction of automatic translation between protobuf and the rest