Hello all, The next Virtual Meeting is scheduled on next Wednesday and will be dedicated to Hardware Topologies Working Group tickets. The WG needs some feedback from the whole Forum and I hope that as much people as possible will be able to attend this next VM. In order to prepare for this VM, I listed a couple of questions to which I hope we will be able to answer. I send you these questions as well as the text for both tickets #132 and #154. If the people that plan to attend could think about these questions and take some time to read the current texts (it's roughly one page each) it would be much appreciated. Questions: Issue #132 (pages 26-27 in doc context-issue-132-10-21.pdf) - Is it OK to *require* an info key (i.e., not just a hint) - In case of errors what would you do? - return a copy of the input communicator handle? OR - return a duplicate of the input communicator ? OR - return is implementation dependent - What is your opinion on having the same outcome for these different cases? - wrong/unrecognized info key value - the output communicator is a duplicate of the input communicator - the info key values provided are different across processes Issue #154: (pages 5-6-7 in doc inquiry-issue-154-10-21.pdf) - Is introducing a new query function the right way to go ? - if yes, which design is preferable? * local version, with the choice of - returning resources to which the calling process is bound - returning resources available and recognized by the MPI implementation * collective version Thanks, Guillaume Thanks! Guillaume