#10 dspsr: predictable archive filenames
If data at the beginning of an observation are flagged as bad, then the MJD of the first time sample folded into the profile may be more than one second later than the start of the data file. In this case, the output archive filename will be >= one second later than might have been expected from the start time of the observation. As this can complicate the collation of results from multiple machines, it would be nice if the archive filename always corresponded to the beginning of the integration, regardless of lost packets or bad data.
Also, when the integration length is less than one second, dspsr defaults to producing an archive filename numbered by the integer pulse turn (e.g. pulse_123456789.ar) ... this is also not necessary. To prevent multiple short integrations with the same filename, yyyydddhhmmss.xxxxxx.ar should be sufficient.
Finally, in single pulse mode, if the pulsar period is greater than one second, dspsr will not produce pulse_xxxxx.ar filenames, and will revert to yyyydddhhmmss.ar It would be better if dspsr always produced pulse_xxxx.ar in single pulse mode (dspsr -s) and always produced yyyydddhhmmss.* in time integration modes (including dspsr -L sec).
