To: new-httpd@apache.org
Subject: Re: cvs commit: apache-devsite how-to-release.html 
In-reply-to: Your message of "Thu, 21 Aug 1997 10:33:46 +0200."
             <199708210833.KAA17370@en1.engelschall.com>
Date: Thu, 21 Aug 1997 16:45:31 -0700
From: "Roy T. Fielding" <fielding@kiwi.ics.uci.edu>
Message-ID: <9708211647.aa24045@paris.ics.uci.edu>
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@apache.org

>>   Remove suggestion to produce a sorted tar file -- there is nothing
>>   professional about it, and it is always less reliable than using the
>>   filesystem directly. 
>
>How do you came across this statement, Roy? Why should be this less reliable?

The software engineering definition of reliability has a lot to do with
probability and Murphy's Law.  Basically, the more hands touch the data,
the more likely that a bug will manifest, but compounding exponentially
instead of linearly because people only test the things which are easiest
to test.  That is why tar of a directory is reliable, tar using append is
less reliable, and tar using append of a sorted list of names goes right
down the tubes.

I say it isn't "professional" to have sorted tar files because I am
professional and I don't sort my tar files. :)

>Although I've no problem doing it the traditional poor man's way, I cannot see
>the your point here. 
>
>>   Use the -exec feature of find instead of an
>>   unreliable pipe.  Use filesystem-based gzip and compress because they
>>   are always more reliable than stdin/stdout processing.  
>
><grin> Can it be that you don't trust Unix pipes? Please Roy, what makes you
>think these things or such problematic?

For the same reason that it took so long for us to get the Apache
buffering code right.  If there is a simpler/faster/more-often-tested
solution available, use it.  Unix pipes are great, but why use them when
you don't need to?

....Roy