Sellam Ismail (vcf(a)siconic.com) wrote: I take it you mean non-volatile clock like the M48T02 RTC chips on the early SUN-4 machines. The reason why there is a difference, is that most operating systems (esp. those that have preemptive schedulers) need a regular timer interrupt. So they use that interrupt for their time-of-day clock. This method is generally accurate and convenient since mainframes were not likely to be switched off overnight. Another advantage held by the early mainframes is that they were kept in a stable environment, which meant that their clock keeping was pretty good anyway. The other reason why you don't actually want to read a time of day clock while the OS is running is speed. Why waste time on I/O since the OS can just keep the elapsed time variable updated with a single instruction. UNIX is a good example: when you boot, the OS goes to the RTC, reads the current time and date and converts it to the UNIX time (seconds since 00:00 Jan 1 1970 UTC). From that moment onwards, the OS is making sure that this value is updated so that it is current. When you actually ask for the time, the system (actually the C library) converts the contents of its time-of-day memory location into the calendar date that humans like to see. When you actually want to set the correct time, the system will update its internal time-of-day memory location *and* the RTC so that then next time the system boots, it will have the correct time. BTW, the original PCs did not have a non-volatile clock, which is why MSDOS by default asks for the date and time every time you boot (assuming you do not override it by creating a AUTOEXEC.BAT file). Since many people mentioned various HP calculators and computers with non-volatile clocks, I only want to mention the HP 59309A HP-IB Digital Clock. This is an HP-IB device that returns the current time/date as an ASCII string. SO if your computer did not have a non-volatile clock you could easily add one. **vp