httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ben Hyde <bh...@pobox.com>
Subject Re: Testing/Validation
Date Thu, 07 Jan 1999 14:43:39 GMT
 
I've a middling complex testing harness here.

I make a sharp distinction between the requirements placed on the
tested client and those on the test driver.  I'm quite the glutton for
rich tools on the test driver while being quite sadistic about the
simplicity of those on the client.

A test driver could reside in the public net as a resource to clients
(apache developers, module writers, porters).  Those clients would
install a test site(s) that attempts to cover a mess of configuration
directives.  The client then requests that test driver study him
providing an account on the test site at that time.

This could start out very light and then hill climb with little risk
of getting too out of control since the test driver wouldn't suffer so
much from the usual installed base problems.

This glosses over that the most difficult part of my testing harness.
That's the stuff to get remote control over the process community on
the client.  There is nothing more irritating than having the entire
testing run fold it's tent just because one machine got one process
holding one port and you've misplaced the PID of that process.  It
happens to me all the time - in part because I've a lot of plumbing
behind Apache.

My number one rule about testing is that unit testing doesn't offer
enough return for your testing dollar to compete with system wide
testing - so always test the whole system first.

I'd love to work on this problem, but I really don't have the
resources.  As a thought provoking subplot consider that
"www.atweb.com" just sold for 200+ million, but I presume that was for
all those locked in web masters they have and not because their tools
are neat (which they are, but not that neat).

 - ben

"the narcissism of small differences."

Mime
View raw message