From mboxrd@z Thu Jan 1 00:00:00 1970 From: tfb@tfeb.org (Tim Bradshaw) Date: Wed, 20 May 2009 09:21:34 +0100 Subject: [TUHS] UNIX turns forty In-Reply-To: <20090519215610.A10612@eskimo.com> References: <20090519215610.A10612@eskimo.com> Message-ID: <3B54F918-4B51-465A-B70D-94A6B72359B6@tfeb.org> On 20 May 2009, at 05:56, Derek Peschel wrote: > Interesting question! And related questions -- When did the current > start of the epoch get chosen? Were there any false starts or early > changes? (I seem to recall reading about one change, moving forward > by a year.) And were there ever any dates in the system that couldn't > be correctly recorded, because the epoch started too late? I'm not sure of the case in very early Unix, but I think in recent (4BSD and later is all I know well) history, time has always been a signed quantity, so you have as long before the epoch as you do after. My wife has an amusing (in retrospect) story about someone who decided it would be interesting to see what happened if you set the clock on a system (these would have been Suns (definitely) running SunOS 4.x (I think, might have been 3)) close to the end of time and see what happens when it wraps: the result was a lot of files with dates in the long distant past, and a lot of work to fix this (which she forced the perpetrator to undertake I think). However I have some memory that really early Unix (a) had a different epoch and (b) counted in different units related to some clock interrupt - 60ths of a second? - which gave a rather short wraparound. --tim