Closed crotwell closed 9 months ago
DATA NEXT 2023-08-09T01:23:45Z 2023-08-09T02:00:00Z
and
DATA ALL 2023-08-09T01:23:45Z 2023-08-09T02:00:00Z
Should initial timed DATA
use ALL
? And have
DATA NEXT 2023-08-09T01:23:45Z 2023-08-09T02:00:00Z
be an error? Or is this possibly useful if end time is greater than current data in server due to latency?
I tend to think that the last variant should be an error. But then we probably don't need the NEXT keyword in the first place.
The '-1' that was renamed to NEXT was initially a dummy placeholder for sequence number in time window requests. We don't need NEXT when time window is not used, because that is the default when sequence number is omitted. And in time window requests we can use ALL.
OK, so can remove NEXT
completely? That cleans things up for the DATA
command considerably. +1 from me
So, possible valid DATA
commands would be:
DATA
- start flow at next packet to arrive at server
DATA 1234
- start flow at packet with seq >= 1234
DATA ALL 2023-08-09T01:23:45Z
- start flow at first packet >= start time
DATA 1234 2023-08-09T01:23:45Z
- start flow at first packet >= start time and seq >= 1234
DATA ALL 2023-08-09T01:23:45Z 2023-08-09T02:00:00Z
- all packets between start and end
DATA 1234 2023-08-09T01:23:45Z 2023-08-09T02:00:00Z
- all packets between start and end and seq >= 1234
Does that cover all cases?
Does that cover all cases?
I think so.
Spec updated accordingly.
How should a user initiate a new connection to get a time window for a channel? The appears that
<seq>
has to be there, but what it should be is not clear from the spec.Should they use:
or
or maybe leave seq out like
or something else?
And related, what should the server do differently for the two above commands? Does -1 "next available" make any sense in a timed DATA command? Seems like you either know the time range you want, or you are resuming and know the start sequence number.
Also see #43 as this may be an additional case where the magic number system is more confusing. Would this be better, using ALL instead of -2 as in #43 ?
Alternative could be to add subcommand words, so initial request would be
and then resume a broken connection with
so seq is optional???
@andres-h what do you prefer for timed data?