W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1997

Re: Origin Servers without Clocks

From: nemo/Joel N. Weber II <devnull@gnu.ai.mit.edu>
Date: Tue, 22 Apr 1997 18:55:40 -0400
Message-Id: <199704222255.SAA01534@diazepam.gnu.ai.mit.edu>
To: Ross_Patterson@ns.reston.vmd.sterling.com
Cc: http-wg@cuckoo.hpl.hp.com
   Date: Tue, 22 Apr 97 18:17:14 EDT
   From: "Ross Patterson" <Ross_Patterson@ns.reston.vmd.sterling.com>

   While I understand and sympathize with the issue here (I've already
   got too many clocks in my home, I don't need more to reset twice a
   year!), I find it suprising that TCP can be implemented on a system that
   has no timing facilities.  For that matter, don't many (most?) embedded
   systems have a real-time kernel, with a timer-based dispatcher?  Or am I
   mistaking a timer for a time-of-day-and-date clock?

Yes, there's a difference between a timer designed to help with mutlitasking
and a timer which tells you the time of day.

Assuming a completely standalone server which has no battery backup,
which boots using DHCP and tftp or whatever (or has all its software
in ROM--though then it still needs at least rarp), there's generally
no need to have a clock to run a soda machine or a thermometer.
Admittedly, a soda machine which logs times people get sodas
is quite interesting.  So is a thermomenter which generates
nice graphs showing time.

Admittedly, if you're going to all this trouble, you could ask
another machine on the network for the time; but I bet there are
other applications where you don't care about the time.

(What about routers?  Some of them can be configured from the web;
but why would a router care about wall clock time?)
Received on Tuesday, 22 April 1997 20:57:00 EDT

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:32:35 EDT