Re: [Gnu-arch-users] [RFC] Naming conventions

[David Allouche]

On Tue, Dec 16, 2003 at 11:09:50AM +0300, Samium Gromoff wrote:
> At Mon, 15 Dec 2003 15:34:12 +0100,
> Thomas Zander wrote:
> > > My assertion: deltas are essential as well, in almost every respect,
> > >   _except_ for the people who don`t use revision control.
> > 
> > Ehh; this does not work with what I thought the dir would contain.  I
> > figured temporary diffs or diffs downloaded from somewhere; but not any
> > originals.
> > Could you explain what it is that you want in the ~/{deltas} dir?
> 
> Products of `tla undo', `tla delta'.

I have skimmed through the discussion so I may be missing something.
But my distinct impression is that you are looking for a problem where
there is no problem.

"tla undo" creates ,,undo-* (remember ,,* junk is special) directories
in the source tree and AFAICT that is the most relevant place to put
this data. I have been chatting with someone asserting that "undo is not
safe". I do not know the precise argument, maybe it is related to the
location of undone changesets, but that seems to be a distinct issue
which needs to be described and discussed in a separate thread.

Other changesets you may create for internal use by a GUI tool are
typically temporary cache data: they can be recreated from a master (the
archive) and are here solely because (1) they are needed at some point
in the operation and (2) they may be used again soon. In that context,
there is no need for anything more invasive to the infrastructure than a
web browser cache.

In a nutshell: leave undone changesets where they are and handle
temporary changesets needed by the GUI as temporary data.

-- 
                                                            -- ddaa