On Wed, Aug 29, 2012 at 10:43 AM, Joost Andrae <[email protected]> wrote: > Hi Rob, > > in my opinion a source control system should not be used to store binary > objects. Quite probable multiple copies of the same binary are stored within > the SVN system if you branch the source tree. Additionally the backup of SVN > will grow as well.
I hear "binary" used in two senses; 1) Files that are built from source files, e.g., binary files are not the original source files; they are derived from source files. 2) Files that are not text files. But there are "original source" binary files. For example graphical files for icons, images, etc. And we already store ODF documents in SVN. Document templates, for example. These are source files, and they are also binary files. SVN is smart about branching, at least on the server. Files that don't change are not copied during a branch. It is a "cheap copy", like a pointer. Of course, if a developer checks out both the trunk and a branch they will get duplicates copie in their working copy. > It's better to have a fixed central place on a webserver to store such > documents. The automation suite could load those documents via http request > in the first run like a well prepared configure script. > But our website is also stored in SVN, right? So we cannot escape that easily... Now, the nice thing about storing test documents in SVN are: 1) Versioning and history 2) Access control allow access to all committers 3) Contributions to SVN are covered by the iCLA. Combined with #2 this ensures they are under Apache License. 4) Auditability -- We can easily determine when a change was made and who made it. Change notices are all sent to our commit list. In theory we could put all the test files on Apache Extras. That would give us #1 above. 2 and 4 would require extra work. And #2 would not automatically happen. -Rob > Just my 2 € Cents... > > Kind regards, Joost >
