comp.os.linux.networking

"Tim Clark" writes:

>In article ,
> ibuprofin@painkiller.example.tld (Moe Trin) writes:
>> On Sun, 16 Mar 2008, in the Usenet newsgroup comp.os.linux.networking, in
>> article , EricThompson wrote:
>.....
>>>It's quite a luxury, and we're paying a price for it now in the
>>>difficulty to which this little discussion alludes. We now have a
>>>world economy to deal with, and travel that within hours takes us to
>>>"other days."
>>
>> International broadcasters seem to have no problem using "GMT" (sic)
>> as do multi-national file systems.
>>

>Exactly: Unix, grabbed the concept very early on. Filesystems and
>computers' clocks need to use an invariant time (let's ignore the leap
>seconds). The problem that any system which uses a civil time which
>jumps back and forwards due to DST will face is that each year there
>will be an hour's worth of times which are missing, and even worse an
>hour's worth of times with an ambiguous meaning. (OK 30 minutes in the
>case of Australia/Lord_Howe and Pacific/Rarotonga, as is revealed later
>in Old Guy's post)

>The solution Unix adopted was simple and elegant: store times in
>UTC/GMT, but display them in the user's civil time as and when required.

>I know MSDOS and the FAT filesystem ran the clock in, and stored, civil
>time. Thus if one had two files on a system in the UK:
> A created at 01:15:00 on October 28th 2007
> B created at 01:40:00 on October 28th 2007
>it would be impossible to say which file is older. It may be that B was
>created first at 00:40:00 UTC when the civil time was 01:40:00 due to
>daylight saving, while A was created later at 01:15:00 UTC when daylight
>saving had ended. Or, of course, they may have been created both in
>daylight saving, or both outside daylight saving, so A is older than B.

Or even worse, at 1:20 AM you ran a program which finds a file created at
1:40AM, obviously in the future. This can drive a filesystem crazy (Ie a
crash)



>Having an operating system and filesystem which can't tell which file is
>older is obviously a major problem. I know NTFS now uses UTC/GMT, but
>are there any remaining issues with XP/Vista + NTFS and timestamps, or
>has it really managed to catch up with Unix/Linux and its simple and
>elegant way of handling the issue?

>Of course some might say it doesn't really matter with operating systems
>where the computer is rarely able to to anything without a user chained
>to it by mouse and keyboard, actually doing the work. Because files are
>rarely created at between 1 and 2 am. But then I wouldn't want to make
>such cheap remarks :-)

>--
>Tim Clark