Will this be true tomorrow?

"Richard Everman" <reverman@ka.reg.uci.edu>
Errors-To: secret@www0.cern.ch
Date: Tue, 8 Feb 1994 22:27:48 --100
Message-id: <9402082124.AA22983@ka.reg.uci.edu>
Errors-To: secret@www0.cern.ch
Reply-To: www-talk@www0.cern.ch
Originator: www-talk@info.cern.ch
Sender: www-talk@www0.cern.ch
Precedence: bulk
From: "Richard Everman"  <reverman@ka.reg.uci.edu>
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Will this be true tomorrow? 
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
Content-Length: 3080
Roy (Sorry Rob)

Is there a basic assumption being made here that the Webmaster will 
always be located at a "local" site; that the Webmaster will never 
telecommute?  What happens if the Webmaster is responsible for UCI, 
UC's Washington DC center, and UC's Education Abroad Program sites 
(which are located in 72 countries)?

Cheers

Richard
reverman@uci.edu

*********************************************************************

From: "Roy T. Fielding" <fielding@simplon.ICS.UCI.EDU>
Date: Tue, 8 Feb 1994 21:11:53 --100
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Re: Will this be true tomorrow? 

Richard said:
> Statement: (hopefully not out of context)
> 
>> Rob said:
>> I prefer local time, myself, as its extremely rare that someone would
>> need to compare stats across time zones 

Actually, I said that, so nobody should blame Rob.

> Question:
> 
> Will this be true tomorrow? next year? 5 years from now?  At the rate 
> the use of the Internet is growing, will we always (at least in the 
> foreseeable future) be interested ONLY in local stats?  If there is a 
> good change that as institutions, companies, etc. become more 
> global (e.g., the USA-based software company that went to a 24 hour 
> help line by switching after hours calls to their AU-based site), 
> shouldn't we also take a global view?  How many microseconds does it 
> take to convert GMT to local time? :-}

Quite a few when you consider how many entries are in a log.  However,
speed was not my primary concern.

Server stats are and will remain a local issue no matter how fast the
Internet grows.  There are several reasons for this:

1. The log file is not directly accessible to the web -- some local processing
   must be performed before it is made available.

2. Machine usage is generally dependent on local users (and their work hours)
   and its generally considered "nicer" to perform large remote accesses
   during that site's off-peak hours.  Thus, as a webmaster, I need to
   evaluate usage at my site in terms of how it effects local usage during
   local peak hours.

3. The log file is intended to assist humans trying to maintain their
   own server, and (except in the UK), all humans think in local time.
   This is particularly notable at the begin/end of each month when the
   log file is truncated by date.

4. Local events (such as power failures, lightning storms, earthquakes,
   due dates for final projects, etc.) occur in local time and, if they
   have some effect on the server, it's easier to match the effect with
   the event if the log reflects local time.

Having said that, it's certainly possible that some people will want to
publish their log in GMT.  However, since I anticipate that to be an
extremely small percentage of web sites, it makes more sense to have
those sites convert their log to GMT (using any number of analyzers)
rather than to require all other sites to convert from GMT to local time.


..Roy Fielding   ICS Grad Student, University of California, Irvine  USA
                   (fielding@ics.uci.edu)