[coyotos-dev] coyotos.sleep, sleepTill, the epoch

[Valerio Bellizzomi]

On 08/02/2007, at 14.29, Jonathan S. Shapiro wrote:

>On Thu, 2007-02-08 at 19:58 +0100, Valerio Bellizzomi wrote:
>> On 07/02/2007, at 21.53, Jonathan S. Shapiro wrote:
>> 
>> >On Thu, 2007-02-08 at 00:06 +0100, Valerio Bellizzomi wrote:
>> >> On 07/02/2007, at 17.37, Jonathan S. Shapiro wrote:
>> >
>> >> >Since we don't report uptime, we don't track it anywhere. There is
no
>> >> >notion of counting uptime at all. I'm confused by the question.
>> >> 
>> >> Why are you confused?  is there any means to know how long the
machine
>> >has
>> >> been up ?
>> >
>> >Nothing in the kernel or the boot mechanism provides any support for
>> >this. There will probably be some mechanism to be notified of
restarts,
>> >and this can be used to implement uptime at user level.
>> 
>> This is no problem. All that is needed to implement uptime is a record
of
>> startup time+date, the counting can be implemented at user level, it as
>> simple as subtracting startup time from current time.
>
>We are converging, but not quite converged.
>
>Recording the startup time and date is not necessary either. When the
>uptime tracker is notified of restart, it simply fetches the time of day
>at that time.
>
>By capturing reboot time in the bootstrap, you are trying to impose
>precision on something inherently imprecise. The only reason to capture
>it that early is marketing numbers, and if your machines are down often
>enough that this small difference helps I think you have other problems.

No I have no problems except my headache :-)

I think we have converged. The fact is that I had forgot about TimeOfDay,
but I found (again) the description in the EROS Object Reference - Kernel
Objects.

http://www.eros-os.org/devel/ObRef/kernel/TimeOfDay.html

So as you say, it is not necessary to record reboot time. I still have
some confusion about how to record the "OS install time" for
administrative purposes, but the uptime question is resolved.


val