Nick Barnes writes:

True. Sometimes though, knowing exactly what the state was isn't the same as being able to reproduce it. Machine configurations themselves aren't always reproducible. There are hardware upgrades, or replacements for defective parts, ditto for networks and routers, and their configurations with other machines, and of course lets not forgot OS patches, disk replacements, and new filesystem maps or mounts. We don't always have every old piece of hardware around (and keeping them around for this sole purpose is often impractical).

Heck, some optimizing compilers that do parallelizing can't guarantee the same bitstream is output for the the same source code and options and environment (though they will guarantee this modulo certain partial ordering changes). So maintaining and tracking what the configuration is/was is often the best we can hope for (cuz we can't always reproduce every one of these conditions).

You could keep, archive, version, track, and control everything under the sun. But at some point you will reach the law of diminishing returns (and when you do may depend on the customers and the application and your market and your balance sheet ;-)

Of course, lots of organization haven't yet reached the point where they feel that doing any configuration mgmt (much less having a CM tool) is "practical" but I hope most of us on this list aren't working at such places.

Sounds very sensible.

--- Brad Appleton <bradapp at enteract.com> | http://www.enteract.com/~bradapp/ "And miles to go before I sleep." | 3700+ WWW links on CS & Sw-Eng