How? It's a derived value. If no server communication, then we'll need to talk to all the owners of datafiles and get their sizes, then do some dist-specific math to get the actual size.
When do we want to ensure that the file size is exact? What does truncate do WRT the datafiles? Does it just make sure that the right one is big enough to show that the file should be so big, or do we do something to all the datafiles?