diff options
author | Eric S. Raymond <esr@thyrsus.com> | 2005-05-20 16:46:24 +0000 |
---|---|---|
committer | Eric S. Raymond <esr@thyrsus.com> | 2005-05-20 16:46:24 +0000 |
commit | 3d3e4e78ac8ecd337e76745b72c7101ffe65eb84 (patch) | |
tree | 3687597f7bc86823cd633eecb75c2198af9fa936 /TODO | |
parent | a777ee947925c8e50727a123c0bad72b6ee0814a (diff) | |
download | gpsd-3d3e4e78ac8ecd337e76745b72c7101ffe65eb84.tar.gz |
Cleanup of dynamic-SHM patch.
Diffstat (limited to 'TODO')
-rw-r--r-- | TODO | 23 |
1 files changed, 0 insertions, 23 deletions
@@ -12,29 +12,6 @@ I don't know if he has checked out SVN head recently. ** To do: -*** Design problems with the NTP interface: - -There's a messy question about how NTP shared-memory segments should -be associated with GPSes. At the moment, the code allocates just -one and all devices report through it. It would be better for each -device to have its own segment, so they look like separate time -sources to NTP and it can apply its heuristics for weeding out -the less consistent ones. - -The trouble with this is that: - -1) The default build of NTP only supports 4 segments. Thus, if we -allocate them per GPS device, we're not unlikely to run out of them. - -2) We may actually want *two* segments per device, one for message time -and one for PPS time. - -3) We can't grab new segments after privileges are dropped. - -I think we're going to be stuck with grabbing pointers to all four -segments just in case gpsd wants them, and then having control-socket -commands to bind NTP unit numbers to devices. - *** Set FIFO threshold to 1 to reduce jitter in serial-message times This can matter if you're using gpsd as a time reference. |