Hi,
WebShield does release the memory used for the connections when the connection is closed (from both sides - that might mean closing the browsing). Not doing that would certainly mean there is a BUG in WebShield but I don't know about such suspicion now. It consumes more memory when working and less memory when idle. Eating 30 or 40 megs is not something I would describe “bam!” or “it hit me again!”
Besides this it also releases its committed bytes back to page file on idle (when there are no connections for several minutes). Not that it would speed anything, help your system or do any good, actually it slows down the process, stresses your disk and make the next web page appear a little slower, but as you can see the value displayed in the "Mem Usage" column in task manager is pretty important for many of you.
Just for comparison I attach an image of my system with XP SP3 running for 22 days (no restarts, just hibernate) with only one tab in Internet Explorer opened as I am writing this message. You can see that IE and WebShield are on opposite sides of the process list (sorted by memory consumption) so I had to split the image in two.
Surely designing an application in a way that it needs as little memory as possible is a good way, but usually when choosing the appropriate algorithm or data structure for some task a developer have to decide between a fast and memory hungry option and between a slower but less memory demanding one. If things wouldn't be so distorted in this whole resource consumption war between users and developers, we might probably be smart and use as much memory as available at any time to speed things up and be more memory conservative (or the system should do that for us - and it actually kind of does) when other applications want memory. Still I thing every byte of memory that is „free“ during some excessive work means my OS is not doing its job properly.
So, please post here if you have a concern that the memory usage for webshield goes only up and never returns back. We might re-do some tests for memory leaks, but I am not aware of any code change that might affect how memory is used in WebShield during recent version updates.
Lukas