??? 08/05/10 22:37 Read: times |
#177746 - Directory copies not really fun for version control Responding to: ???'s previous message |
I use CVS or similar for just about everything.
Source code, configuration files for servers, todo lists, ... It's a lot of extra work to fool around with backup directories. With a real repository, I can request an annotated list of a file, where every single source line mentions which time and owner of commit. And when viewing the difference in a file between file revision 1.22 and 1.23, I can see my comments I wrote at the time when I commetted the changes - for example: "Bugzilla #917: Added support for redundant configuration servers." Between two releases, a single file may have changed 20 times. I then have all these 20 copies available, together with 20 comments regarding the changes. An I can instantly select between which two versions I want to see a diff. Since I do a lot of Linux development, I'm quite comfortable with setting up or configure machines so it doesn't take many minutes to get one more repository including working backup. But as I mentioned earlier: it's possible to download a complete virtual machine with CVS or SVN preinstalled. And one physical machine can handle many virtual machines - especially since a virtual machine only used for a repository needs no CPU time and very little RAM. There are many alternatives for backup: - taking a copy of the full virtual machine image. - letting the virtual machine share a directory tree as SMB or NFS, to let some machine on the network backup just relevant files. - letting the virtual machine share a directory tree with the host machine. - letting the virtual machine run a backup program that sends backups to a shared disk or backup server somewhere. For a company that already have working backup solutions, it doesn't take special skills to have such a virtual machine fully activated with backup in 1-3 hours of work. Obviously less, if the company do have someone skilled. But going the zip-file or directory-tree-duplicate route is a very inefficient method. It is prone to accidents or forgetfulnes while at the same time giving much worse trackability. A repository is not just great/required/obligatory for team projects. They are just as useful for one-man operations. I think I have even seen (but alas don't remember where) some cheap Ethernet-connected disk box that contained not just a web server and ftp but also CVS and SVN. You added two internal disks for mirroring. Powered it up and gave it a name and IP. Then either configured backup to an USB disk or a shared disk volume somewhere on the network. |