The nice thing about having files in the database is that you can have constraints set up to keep files from being orphaned. For example, let's imagine we were implementing sourceforge, in perl and postgres. Sourceforge needs to deal with a lot of files, with projects and files being created and deleted all the time. If everything's in the database, if you delete a project, you can set things up so that a cascaded delete will remove all the files from the database at the same time, and you can tell, with a simple query, if you don't do that, which files are orphaned. Moreover, your database backups, if done right (check the pg_dump manpage), will include the appropriate version of all the files in there, so you'll never need to worry about them getting out of sync with your regular backups. Finally, you don't need to worry about a separate access path to manage the files referenced in your database -- if you can SQL in, you can upload/update/delete the files (subject to database ACLs, of course), and there's no need to also allow ssh, ftp, or whatever other solutions to manage this stuff. In sum, it's easier to get all the smarts that your database has to cover files if the files are in the database.