Re: Space wasting

Adal Chiriliuc wrote:
> I've studied the amount of space used for exporting or checking-out
> the Boost library so that we can compare how much overhead SVN adds.
[...]

A Subversion working copy does indeed require at least a 100% space overhead by keeping pristine copy of each file, and a large overhead in the number of files (about 3x to 4x) and directories (about 10x).

Some people have done some work towards compressing the pristine copy (also known as the "text base") or optionally eliminating it (for situations where the network and server are fast). I think one or both of these options will probably be in a medium-term future release of Subversion.

The overhead in the number of files and directories is not currently being addressed, as far as I know.

> I'm curious what's the situation on Linux filesystems.

More variation. There are several file systems used on Linux, and some of them are much better at handling many small files than the ones you quote, and some of them are probably worse.

> Why doesn't SVN use a single folder? Why does it need 9 subfolders?

It's a clean design that works well and is not a big overhead in most situations. The number of folders used was one of the least important things to worry about. Now that people are sometimes encountering situations where the number of folders is a significant issue, there is a possibility that someone may propose a modification to reduce the overhead. However, note that the space overhead of these folders is only really significant in projects that average less than about three files per folder, and whose files are all small, which is quite unusual. Most of the overhead typically comes not from the folders but from the three or four extra files per project file.

> Also, is the README.txt file or the empty-file really needed?

No, they are not really needed, and they might be removed in future, but nor are they a big part of the overhead, because there is only one of each per project directory, compared to the three or so extra files per project file.

Anyone wishing to help reduce these overheads should concentrate on the biggest contributors, which are the text base (most significant when files are large) and the properties files (most significant when files are small).

In the meantime, we have to accept that a Subversion working copy requires at least twice the space and sometimes several times the space (depending on file system and distribution of file sizes) that the versioned files would take up if unversioned.

- Julian

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Mar 8 15:44:31 2004