Timezones

luotonen@ptsun00.cern.ch (Ari Luotonen)
Errors-To: secret@www0.cern.ch
Date: Wed, 9 Feb 1994 14:58:59 --100
Message-id: <9402091356.AA14343@ptsun03.cern.ch>
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: luotonen@ptsun00.cern.ch (Ari Luotonen)
To: Multiple recipients of list <www-talk@www0.cern.ch>
Subject: Timezones
X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
Content-Length: 1815

Clearly we cannot make a decision of weather everybody wants
localtime or GMT.  I'll make localtime default, and provide a
-gmt option to allow times in GMT.  Everyone should be happy
with that.  If a given statistics collecter requires a certain
timezone then use that one.

But I do need to know exactly how the timezone info should appear
in log file (if it is necessary):

	1) not at all -- WebAdmin knows it anyway
	2) as GMT +/-NN   or just  +/-NN since GMT always appears in
	   there. Is ftime() portable?
	3) a 3-letter abbreviation for timezone -- I understood that
	   they are not standardised;  how would I get that in C
	   in a portable way.  Is strftime() portable?

My personal comments:

Giving an option is better than restricting to one fixed format
which doesn't satisfy nearly everyone.  This was also what I had
in mind in my original message but I was obviously misunderstood.

If I am only interested in how many times a document X is accessed,
I do _not_ want to have megabytes of diskspace wasted in access
times (whatever timezone), hostnames, remote idents etc.  Since I
want my own kind of special logging then I don't need nor do I want
to run a statistics collector, I just say

	sort log | uniq -q | sort -r > my_stats

and I am very happy.  If I want to use a standard statistics collector
fine, then I just use the standard format.  I thought having fully
configurable log files was a good idea, but some people just fail to
see that the options do not have to be used if they happen cause
nightmares.

IF you want to use them fine, you can have whatever logfile format.
BUT it may then not be understood by statistics collector, which you
probably won't be running in this case anyway.

[I'm inspired by the 31MB log file on info.cern.ch from these 8 days.]

-- Cheers, Ari --