[LEAPSECS] Crunching Bulletin B numbers (POSIX time)
Ian Batten
igb at batten.eu.org
Mon Feb 21 04:03:33 EST 2011
On 21 Feb 2011, at 03:32, Paul Sheer wrote:
>>
>> [...] But so what? It'll report 00:00:03 Monday when it's "really"
>> 23:59:57 Sunday --- why is this small number of seconds any more
>> important than any other small number of seconds error? [...]
>>
>
>
> A miss is as good as a mile when a timestamp is an index into a hash
> table.
So let's try to understand your use-case. Machine A, which is isolated from the outside world and unable to perform leap-second adjustment but maintains a stable clock with a CsRb , is generating hashes, alongside Machine B, which is performing leap-second adjustment via (for the sake of argument) NTP or GPS., and those hashes are being used as indexes into a hash table that both A and B read and write? Yes, I can see leap-second would cause a problem there, except (a) the engineering design is insane and (b) even if it isn't, TAI is ready and waiting to solve their needs. Has anyone, anywhere, written a system so stupid and not been locked up?
>
> These systems don't care whether the event really did happen at
> 12:34:56pm nor whether the clock was a few minutes slow that day.
>
> But they DO care that they are talking about the SAME event!!!
>
So how is a leap second (or, more to the point, 120 leap seconds) any different to "a few minutes slow"? Two systems are communicating. One applies a leap second the other doesn't know about via nasty slewing over the course of an hour. Two other systems are communicating. One loses a second relative to the other through thermal effects. Describe how an outside observer might distinguish those scenarios given a sequence of timestamps from the systems.
>
> Imagine a complex business environment represented by a large block
> diagram. Data, including fields that contain timestamps, propagate
> around this block diagram -
A complex business environment which nonetheless cannot perform one bit of I/O per eighteen months. A complex business system which nonetheless cannot interact with NTP, MSF/DCF77/WWV or GPS? I think you're making something up to bolster a point, not referring to any real business environment.
ian
More information about the LEAPSECS
mailing list