Daniel Popowich
dpopowich at comcast.net
Wed Aug 11 12:42:56 EDT 2004
> I would thus have though saying: > > if not klass or mtime != cached_mtime: > > would be more appropriate. Ie., if the mtime is different to cached > mtime and not just newer. Hmmmm. Thought provoking. Of course, what's really behind any logic like this is "has the code changed since last it was cached?" The modification time is the canonical way to check because looking up inode attributes is much faster than any other method and is "good enough" (yes, the mod time could change without any of the content actually changing, like via touch, but we're willing to accept that because any other method of checking would be just as expensive, if not more so, than exec'ing the source). While I don't have a problem with changing this in mpservlets I have this nagging feeling there's a good reason for keeping looking for future modification times as opposed to any modification time. However, non leap to mind. Can anyone think of any reason not to go for this change? Daniel Popowich ----------------------------------------------- http://home.comcast.net/~d.popowich/mpservlets/
|