directories size is always 4K
 

Why is the directories size always 4K ?

milosevic@slackws:~$ ls -ldh /
drwx--x--x 25 milosevic users 4.0K 2005-12-10 19:47 //

I mean, shouldn't it be the sum of the sizes of the files that the dir contains? Or does it have anyhing to do with the inode's size?

Cya fellas.

I can't tell you why your root directory is 4K in size,
but the fact that it's not owned by root:root is a worry.

Who set-up that box with those incredibly stupid ownerships?



Cheers,
Tink

hm... sorry... actually the thing is like this:

milosevic@slackws:~$ ls -lhd /
drwxr-xr-x 19 root root 4.0K 2005-12-09 15:22 //

it's the block size of the filesystem. the size of a directory and the size of all files within it are a totally different thing.

Hmmm ... nice attempt of an explanation, but I don't think
it's correct. On 4 different machines here I get four different
sizes, and they're all under 1K (616,624,704,512) ... so the
"human-readable" didn't quite mess up the results.


Cheers,
Tink

If you're trying to find out how much space is taken up by say, root (/) and its subdirectories, you will want to use the du (disk usage) command:

As far as the listed size of the directories is concerned, let's experiment, watching the directory size of /foo grow as we progress:

Now, without checking the source for ReiserFS or ls, one can make an educated guess that what is reported by the directory size is the size of the directory itself - that is, the size of the directory name and its contents' names and attributes.

Believe it or not, each directory entry you make takes up disk space - more on some filesystems than on others. Take FAT for example: if you create a new, empty directory that directory takes up space in the file allocation table. If the filename is longer than a FAT-16 8.3 filename, behind the scenes that directory is broken up into multiple directory entries (a FUGLY hack, I might add) and that all takes up at LEAST one cluster (anywhere from 512bytes to 32KB depending on formatting options). EXT2 works in a similar way - each directory takes up a cluster, plus there needs to be meta-data stored, that is, the directory's timestamp, permissions, etc. not to mention the mappings to the files that directory contains. Some other filesystems are a little more efficient and store this info in a b-tree structure (ReiserFS, XFS, and if I recall correctly UFS as well) which resmbles an RDBMS more closely than a conventional filesystem.

I hope this helped shed some light on things rather than confused you more. Filesystems are actually fairly complex structures.

file size.
 

If every file takes up at least one cluster of the file system, why the hack the "file.txt" only takes one byte? (i believe the cluster size here is 4k).

Cheers, Milosevic.

Well some filesystems (ReiserFS) are "zero-slack" which means that if a file is 1 byte, it takes one byte (well, plus a couple of delimiters). Others (ext2, uncompressed NTFS, FAT) require that whether file is 1 byte or 511 bytes on a filesystem partitioned with 512-byte clusters, that the file be allocated a minimum of one FULL cluster - that is, 512 bytes. The wasted space is usually referred to as "slack"

Why does LS report the file size as one byte? Because the file's contents is one byte. You're confusing file size with disk usage. If you dual boot Windows, here's a quick demonstration:

On an uncompressed FAT or NTFS partition, create an uncompressed file which is say, *picks an abitrary number* 42 ;) bytes. Open up that file's properties, and in the properties tab, check out two entries:

* file size
* space used on disk (I forgot exactly what the two fields are called)

The numbers will be different. One will show the size of the file's contents, and the other will show how much space on the disk is allocated to it. The difference between the two is allocated but unusable space called "slack"