Devuan fork of gpsd
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

302 lines
12 KiB

  1. This is the gpsd to-do list. If you're viewing it with Emacs, try
  2. doing Ctl-C Ctl-t and browsing through the outline headers. Ctl-C Ctl-a
  3. will unfold them again.
  4. For contribution guidelines and internals documentation, please see
  5. the "Hacking Guide" on the project website.
  6. The list of bugs exposed by gpsd in other software has moved to
  7. the "Upstream Bugs" page on the project website.
  8. ** Installation **
  9. Improve the uninstall target. Remove library syminks.
  10. Make libdir conform to new FHS layout: 32bit in lib/ and 64bit in
  11. lib64/. Doing that also makes the uninstall more complex.
  12. ** Documentation **
  13. We now have two different calibration procedures in the Time Service
  14. HOWTO. One was written by Gary and is based on peerstats; the other by
  15. Sanjeev Gupta and is based on loopstats.
  16. Can these be merged? How do we know which is appropriate to use when?
  17. What could we build in the way of tools and visualization aids to make
  18. the calibration process less annoying?
  19. ** Things to do when we can break compatibility **
  20. -------------------------------------------------------------------------
  21. In gps_data_t, save PPS precision; this will require creating a pps struct.
  22. Where would PPS precision come from?
  23. Make the Python JSON client as smart as the C JSON client. Gonna be
  24. a big job. The C client checks all the JSON classes for completeness
  25. and correctness. For example, a missing JSON field will be replaced
  26. with a default value. Also out of bounds values are fixed. The Python
  27. client does no checks. it just mushes a JSON sentance into a big Python
  28. variable. No validity checking, no missing fields fixed, etc. The C
  29. client uses a generic validator function, fed with templates for each
  30. JSON message. Initially harder than a simple procedural parser, Once
  31. running it is very robust and easy to tweak.
  32. Add u-blox IMU (UBX-ESF-RAW) support. Nothing we can do without
  33. samples or a u-blox IMU.
  34. Adding a sample Go client (gpspipe -> pytogo -> Go ?).
  35. Move out of contrib. It needs a man page and light cleanup.
  36. ** Bugs in gpsd and its clients:
  37. *** Tracker bugs
  38. See the GPSD bug tracker on the project website, but don't be
  39. surprised if it's empty or very sparse. Our rate of new defects per
  40. month is quite low.
  41. *** RFC2783 bugs
  42. **** detangle TIOCMIWAIT and RFC2783
  43. Currently, the code for RFC2783 seems to depend (logically) on
  44. TIOCMIWAIT support. However, RFC2783 (and really, no standard)
  45. defines TIOCMIWAIT, and NetBSD (and probably FreeBSD) do not implement
  46. it. So, there really should be two separate implementations, one that
  47. works with TIOCMIWAIT (and thus only on Linux and OpenBSD) and one
  48. that assumes only RFC2783 (and thus works on Linux, FreeBSD and
  49. NetBSD, as well as any future system that complies with RFC2783 and
  50. uses the serial port for time_pps_init).
  51. *** Client bugs
  52. *** Dispatcher/network issues
  53. **** Reading AISHub data via UDP confuses xgps with short writes
  54. To reproduce, "gpsd -D 5 -N -n udp://" (only
  55. works inside and run xgps. Probably some kind of
  56. packet aggregation issue as it doesn't happen with test logs.
  57. *** Driver issues
  58. **** RTCM3 analysis is incomplete
  59. We can presently decode RTCM3 messages of type 1001-1014, 1029, and
  60. 1033. This is not the complete set. We need more test data with
  61. different types in them, and a copy of the RTCM3 standard at latest
  62. revision (all we have is revision 3).
  63. The support for unpacking RTCM3 sentences in the client library is
  64. limited to 1001, 1002, 1007, 1008, 1009, 1010, 1014, and 1033. There
  65. are some design issues with the baby JSON parser that are going to
  66. make extending this set difficult.
  67. **** Reporting code for specialized Type 6 and 8 AIS messages is incomplete.
  68. This one is mine and Kurt Schwehr's. Support is currently nearly
  69. complete; the only missing cases are a handful of IMO 236 and IMO 289
  70. message 6 and 8 subtypes - specifically, 6/23, 8/21, 8/22, 8/24, and
  71. 8/26.
  72. ** New features **
  73. ** Qt **
  74. Add a sample client
  75. *** Time tracking
  76. Our goal is to be independent of the system clock.
  77. Create a CLI tool to convert UTC to GPS week/tow and back.
  78. **** Maybe apply wraparound compensation ****
  79. See <>
  80. *** Optionally use tag blocks in NMEA output
  81. Ed W. writes:
  82. >Feature request: I would like to see an option to get the NMEA data
  83. >out of gpsd, but augmented with a timestamp on each sentence and
  84. >details of which input fed in the data. TAG blocks appear to solve
  85. >that problem perfectly. So the feature request is simply an option
  86. >that inserts a TAG block for each emitted sentence from GPSd with a
  87. >timestamp and some handle which references the input method for the
  88. >sentence.
  89. >
  90. >I think all this is effectively emitted with the JSON output (?), so
  91. >really this is just adding it to the NMEA sentence output options via
  92. >the TAG structure
  93. However, Iván Sánchez Ortega notes:
  94. >There is one *big* problem with using TAG blocks to timestamp the sentences.
  95. >The standards (as far as I/we know) say that the timestamp is **either** the
  96. >number of seconds **or** milliseconds since the UNIX epoch, but we do **not**
  97. >know if the initial state should be secs, msecs, or nothing. We do not know
  98. >how clients reading that TAGblocked NMEA stream should ask GPSD to switch the
  99. >timestamp units (or enable/disable them).
  100. >
  101. >There are a few sentences introduced in NMEA 4.00 that supposedly allow that -
  102. >the client sends a sentence to the server, and then the server starts
  103. >TAGblocking sentences. But, alas, there are no examples anywhere.
  104. >
  105. >>From my personal experience, the way to know if a TAGblocked timestamp is in
  106. >seconds or milliseconds is ask the people in charge of setting up the
  107. >device/service. Perhaps some devices output only secs/msecs, and some other
  108. >accept the NMEA 4.00 TAGblock configuration sentences.
  109. >
  110. >I, for one, would like to see the secs/msecs problem solved before GPSD
  111. >embarks on the enterprise of writing TAGblocks.
  112. If someone is working on this, please see:
  114. **** Speed, mode and rate-changes in client-mode gpsctl.
  115. Baud rate and mode changes work in direct mode but are not
  116. reliable in client mode.
  117. **** Optional dump of GPS configuration using gpsctl.
  118. Some settings, like antenna in use, dead-reckoning mode, would be
  119. helpful to see.
  120. **** Integrate 1PPS into profiling ***
  121. We caw now *really* measure latency from GPS top of second when it has
  122. 1PPS. Add this capability to gpsprof and revise the "Where's the
  123. Latency?" white paper.
  124. *** Low-power autoconfiguration in the uBlox driver ***
  125. Anthony Stirk <> writes on Wed May 21 06:09:00 2014:
  126. The low power modes are really good on the ublox modules (especially the
  127. MAX7's) however a word of warning.
  128. We use the 1 sec cyclic mode on our balloon trackers but if the module
  129. looses satellites or is jammed out (cheap Chinese cameras do it) the ublox
  130. modules seem to hang and become unresponsive. To mitigate against this you
  131. need to put the module back in max performance mode as soon as the
  132. satellites drops below 4.
  133. I'm not sure for most applications the low power mode needs to be used at
  134. all on the newer ublox modules. The values for the MAX-7Q are 22mA (6Q
  135. 50mA) under acquire, tracking (i.e locked and in max performance) current
  136. at is 17mA (39mA 6Q) and power saving reduces this to 5mA or less (15mA on
  137. the 6Q). We've observed the issue on the 6 and 7 series.
  138. A possibility: drop the chip to low-power mode when it can see > 4
  139. sats, revert to normal otherwise.
  140. *** Make subframe reports available in the C API.
  141. gpsd now reports subframes when they're available, but the C client
  142. library doesn't yet parse them. A sample program to dump the SUBFRAME
  143. data would also be useful. Gary Miller has this in his queue.
  144. *** Write advanced features for TNT Revolution device
  145. The baud-rate switcher in the TNT driver needs to be tested.
  146. gpsmon could support a number of TNT configuration commands, including
  147. unit changes. See
  148. for possibilities.
  149. Jon Schlueter has one of these on a flock machine, so testing
  150. shouldn't be difficult.
  151. *** Enable flocktest on the Debian server farm
  152. Debian server farm boxes have a screwy chrooted environment setup.
  153. flocktest needs to be modified to deal with it.
  154. *** Finish gpssim
  155. It's blocked on skyview computation.
  156. *** Complete and test the speed/parity/stopbit methods in the drivers
  157. These are used for the '?DEVICE' command. All work for 8N1.
  158. **** superstar2: not implemented (driver is unfinished)
  159. **** italk: not implemented (but could be)
  160. **** tsip: speed tested, parity and stop-bit switching not tested
  161. **** sirf: speed tested, parity and stop-bit switching not tested
  162. **** evermore: speed tested, rejects parity/stopbit setting
  163. **** ubx: fully implemented, parity and stop-bit switching not tested
  164. **** zodiac: fully implemented, not tested
  165. **** navcom.c: speed tested, rejects parity/stopbit setting
  166. SiRF, UBX, TSIP, and even Zodiac can support non-8N1 modes; these need
  167. to be tested.
  168. *** Per-driver restore of changed config settings on exit.
  169. This is a solved problem for generic NMEA, EverMore, TripMate,
  170. EarthMate, TNT, Zodiac, and RTCM104 drivers (if only because they
  171. don't configure any device settings).
  172. The SiRF driver now restores NMEA when necessary. It also restores
  173. some (but not all) of the things it tweaks in binary mode -- at the
  174. moment, just the Navigation Parameters from message 0x13. With more
  175. work, we should be able to do a full revert.
  176. The TSIP driver changes its per-cycle sentence inventory and thus
  177. needs some state-restore logic. This can be done; the same packet 0x35
  178. we use to configure it can be sent in a no-argument mode to query
  179. the current sentence mix for later restore.
  180. The FV18 changes its per-cycle sentence inventory to include GSAs. It
  181. is possible to query that inventory, though we don't have code to do
  182. it yet.
  183. Garmin devices are a mess. We reconfigure those heavily, and we
  184. don't know if there's any way to capture their configuration state
  185. before we do it.
  186. The iTrax02 driver sets SYNCMODE to start navigation messages without
  187. checking to see if it's already on, and stops navigation methods on
  188. wrapup. It also forces the set of sentences issued. There doesn't
  189. seem to be any way to query these settings.
  190. ** Future features (?)
  191. *** Industry-standard format dumping of raw satellite data
  192. It would be useful to be able to report raw GPS data (pseudoranges,
  193. clock, doppler carrier) from any GPS device that can report
  194. pseudoranges etc. This belongs in the daemon because the device
  195. drivers are already doing the packet-cracking needed to get the data
  196. off the chips.
  197. Several commodity chipsets (ANTARIS, iTrax3, SiRF and Trimble) readily
  198. output enough data to make this a chore, rather than a hard problem.
  199. Probably the best way to do this would be to add support for unscaled
  200. output of pseudoranges in SKY and add clock and doppler carrier
  201. phase. This JSON could then be interpreted by a client program and
  202. dumped in various standard formats.
  203. Currently the RT-IGS format is looking like the favorite for
  204. implementation; it's a fairly lightweight protocol, flexible enough to
  205. handle all the quantities required, and it is actually in use in
  206. production reference networks. RT-IGS is also a packet-oriented
  207. format, rather than a file-oriented format like RINEX.
  208. *** Support for more survey / professional / up-scale receivers.
  209. Devices such as the Javad JNSCore, Hemisphere Crescent, Septentrio
  210. AsteRx and PolaRx, NovAtel Superstar2 and OEMV, Thales (Magellan
  211. Professional) AC12 and DG14 would all be welcome. Of course, these
  212. are not $50 USB mice...
  213. Local variables:
  214. mode: outline
  215. paragraph-separate: "[ ]*$"
  216. end: