In terms of what to do with writing the output back over the input, I think it would help if you can support version control semantics where possible, e.g. WebDAV for HTTP, or perhaps support for common version control systems for filesystems. For databases and content management systems, presumably they have their own versioning mechanisms. You do need to be able to do a genuine non-versioned overwrite for some production systems that have had sufficient testing, but otherwise I think versioning (beyond just a single ".bak" file) is the way to go. I still remember how handy it was that VMS did file versioning by default when you saved a new copy of a file. Cheers, Tony.