[LEAPSECS] a modest proposal

[Joseph S. Myers]

On Mon, 11 Feb 2008, Steve Allen wrote:


> On Mon 2008-02-11T13:31:17 +0000, Tony Finch hath writ:

> > How should the kernel interpret time stamps stored in filesystems? Do you

> > propose to retrospectively re-interpret them as being in TI instead of

> > POSIX time? (This is related to the problem that Unixes have with FAT

> > filesystems that store timestamps in some unspecified local time, which

> > implies that the kernel can't be ignorant of local time.)

> 

> I suggest that the question of retrospective time stamps is not

> interesting, for it is rarely possible to be sure that a system clock

> was set "correctly", and it is never possible to convince history to

> revise its notion of what time it thought it was (today is George

> Washington's birthday, O.S.).


The question does not relate to the *correctness* of those timestamps, 
it's a case where consistency is much more important than correctness.

If a system has a timestamp and exports an interpretation of that 
timestamp, and another system stores that interpretation or a second 
timestamp derived from it, and then one system changes its interpretation, 
the timestamps are no longer in sync between the two systems.  This can 
cause issues such as unnecessary remirroring of very large amounts of data 
whose timestamps no longer match at the two ends.  See
http://lists.debian.org/debian-user/1997/03/msg00075.html
for an example of this from when Debian made the mistake of trying to go 
it alone with a different interpretation of timestamps.  There are further 
discussions in the debian-devel archives from March-May 1997 of things 
that broke.

If you do not change the timestamp interpretation formula in POSIX, but 
stop inserting leap seconds at some point (or define that timestamps after 
that point are in TI not UTC), you can avoid the issue of reinterpreting 
existing timestamps.

If time signals start giving TI and so de facto civil time becomes based 
on TI without associated legal changes, of course you bring back 19th 
century legal problems and enrich the lawyers.  We know that a nine second 
difference (between TI used by clocks and GMT as a basis for legal time, 
say) is sufficient to be legally significant: 
http://www.theregister.co.uk/2007/11/07/delay_sinks_unfair_dismissal_claim/

-- 
Joseph S. Myers
jsm at polyomino.org.uk