The sizes of files and directories that are copied from one directory to another will almost always be different! But that's OK since the reason relates to how data is stored.
When you create a new directory, it has very few slots to hold names so it must be expanded to hold new names. If a directory is filled with hundreds or thousands of names, then most of the names are removed, the directory will be shown as having a large size (with ls -ld) but only a few files--the directory does not return the empty name slots. Thus, a copy of a directory will be significantly smaller because the unneeded slots are not created. But the copy is just fine even though the occupied amount of space is smaller.
On the other hand, copying files may result in a larger copy. This is due to sparse files. All Unix system can be directed to create a sparse file by writing an inital record, then, using lseek(2), writing record number 1 million. The files contains 2 records with a million imaginary spaces in between so the file is small on the disk.
But copy this file serially (it was never created serially) and all the missing (undefined) records are supplied as a stream of zeros, thus the copy is significantly larger. The most common sparse file is a core dump which can have empty spaces where gaps in the memory allocation exist. Now the file is identical to any program that reads the data whether it is the original or larger copy--they produce the same exact records when read by an application.
Note also that cksum is a good way to verify files and will produce the same result for sparse files and their copies.
Bill Hassell, sysadmin
