1. A user selects a group of files to download as a single zip and then
clicks the download link.
2. The server creates the Zip file (I've confirmed that the memory used by
the zip creation process is being released).
3. The server then streams the zip file from a temp directory to the
client.
4. The user is immediately prompted for the location and filename to save
the zip file as.  However, during this time the memory utilization on the
server increases by 5 times the size of the zip file (i.e. a 10 MB zip
file consumes 50 MB of memory during this transfer).
5. Until the entire file is streamed to memory on the server, IE on the
Mac is displaying a status of "Waiting for data".  Once the file is
streamed to memory, IE on the Mac begins the actual download.
6. Once the download is completed, the memory on the server does not get
released (I've let the server sit for an hour with no activity to see if
the GC would do the cleanup, but it did not).

(编辑:aniston)