...
- outcome of the sync meeting with Ulf from the CCS project
- Stephan: we could change the VSS database for SQLite and request for the properties on demand
- Gunnar: there are 3 cases that can be considered
Do we consider relying not only on cached/stored/updated values in database updated periodically by vss feeder or foresee the need of “on demand” requesting the fresh data ? These 3 ways of accesses are not considered as VSS leaves definition, but might be specified using VSS Layers. Then VSS feeder/Data Server (discussion on the role) can act accordingly (use cached or request for fresh value).
- TODO Gunnar to write prepare a block diagram describing the various cases
- Alex :agrees with the 3 cases
- Alex: do we really need a database which might be seen as an overkill for me ? some hash-map in memory might be enough
- Alex: when the data are not changed at all, what do we do ?
- Gunnar: do we need a database ? no, but there is an implementation that exists
- Alex: are the feeder and the grapql in the same process ?
- Gunnar: I think there are separate processes and we need IPC soemwhere unless we put everything in one process
- Stefan: but even if we use one processor we need some thread sync.
- Piotr: IMHO using hashmap and sync implementation would result in the same complerxity than using SQLite
- Gunnar agreed
- Gunnar: debouncing ?
- Gunnar: there might be some higher performance database, but basically we need only a key value store
...