I have been doing a survey of clock accuracy
(http://vancouver-webpages.com/time/) and have found that, while the
majority of Web server clocks (as derived from the http Date header)
are very accurate, a few are way off.
I then wondered what the effect was of running an http accelerator (Squid
in httpd_accel mode, for instance) and tried it. It seems that the
Server header is unmodified, but that the Date header remains set at
whatever the origin server said until someone does an unconditional GET
(shift-reload in Netscape) (or presumeably until the original page is
modified...).
I can't tell (as far as I know) if this is actually the reason for
seeing clock errors.
I wondered if this was the way that httpd accelerators were intended to
work.
Andrew Daviel
Vancouver Webpages & TRIUMF
Received on Sun Mar 08 1998 - 13:52:22 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:39:12 MST