TMG-L Archives

Archiver > TMG > 2001-09 > 0999404613


From: Eric Haas <>
Subject: Re[2]: [TMG] Julian dates
Date: Sun, 2 Sep 2001 00:23:36 -0400
References: <004601c13326$c7ac0d80$b64279a5@oemcomputer><3B9163EE.FCE83B78@infoave.net> <001901c1334c$2a1d9f40$574a79a5@oemcomputer><010801c13365$a801faa0$c7405041@hpp900>
In-Reply-To: <010801c13365$a801faa0$c7405041@hpp900>



Sunday, September 02, 2001, 12:06:28 AM, you wrote:

WMR> Seems like when I was designing systems we had a "Julian" date which was a
WMR> combination of the year and the numeric day of the year. Since we were only
WMR> interested in dates in the 20th century at that time (Y2K problem) we used
WMR> the last two digits of the year and the numeric day of the year such as
WMR> 75001 for 1 Jan 1975. This was done in order to store data in a minimum
WMR> amount of disk space. The largest numeric value we would encounter with this
WMR> system was 99365 which we could store in 17 bits (yes we used the number of
WMR> bits required for the data). Since we were not interested in dates prior to
WMR> around 1950, we created an algorithm to pack the date into 16 bits.
WMR> We did not have the luxury of many gigabytes of storage like today and it
WMR> was almost a necessity to do this kind of manipulation in order to store
WMR> data that had many many dates that items were due.

Why not just count days from Jan 1, 1950? That would have allowed you to
specify any date until roughly August 2039 in 15 bits.

This thread: