[ jEdit-users ] Internal version control system & getting rid of Save

[Kristian O. <kri...@he...>]

I'm taking a course on GUIs at my university, and I got an interesting
idea that gets rid of the "Save" feature completely by saving the file
automatically when needed. To make this work properly, the editor
needs an internal version control system, much like CVS but simpler,
that lets the user go back to any earlier version of the file. From
the users' point of view, the file on hard disk and the editable copy
in memory are always synchronized, so he doesn't need to think about
saving at all. Also, he gets the benefits of version control.

A local GUI design/research group is doing research on how to do this
in MS Word. The research is still unpublished (and unfinished, I
think), but the lecturer, who is part of the research group, told the
main idea. I began to wonder if this could be usable in jEdit.

Here's how it works: jEdit periodically checks if the file has been
modified since last auto-save. If so, it generates a diff between the
versions and saves it in a file that contains the version history of
the edited file. This is much like CVS. The edited file itself is also
saved, in a normal (plain) format that makes exporting easy. The whole
thing is transparent to the user.

IVCS also supports snapshots, or tags in CVS language. The user
creates a snapshot of his work whenever he wants, and can easily go
back to it whenever he wants. He can also go back to any version, as
in CVS.

As jEdit is a programmers' editor, there should be support for working
with projects, like making a snapshot of a bunch of files. However,
IVCS is not ment to "replace" CVS, but to make text editing more
comfortable.

There is a project called Subversion that is supposed to replace CVS
eventually. The Subversion project documentation has a good
description of a modern version control system. I think it's worth
reading before designing IVCS.
The URL is <http://subversion.tigris.org/index.html>.

This would require quite a bit coding (but at least the diff algorithm
is implemented in jDiff:) and there are a lot of details to be worked
out, but do you think this is a good idea?

-ko