Re: [Monotone-devel] Commit without working copy

[Grzegorz Jakacki]

On 12/2/2004, "Nathan Myers" <address@hidden> wrote:

> On Thu, Dec 02, 2004 at 11:37:26PM +0800, Grzegorz Jakacki wrote:
> > When a webpage is requested, my webserver runs PHP script. I would like
> > this script to run 'monotone cat' on database stored locally on
> > webserver's host, capture the output and serve it as a webpage to web
> > client. At this point user can manipulate the served content in its web
> > client, and then send it back to the web server. The web server should
> > in turn run monotone and commit the new version of the content to the
> > monotone database (stored locally on webserver's host).
> >
> > So basically I would like to get revisions in and out of monotone
> > without creating a working copy.
>
> This sounds equivalent to "I would like to be able to commit individual
> files".  

Not really equivalent. I would like to be able to checkout and commit
files without writting anything to disk.

> Is this just because you don't want to waste the disk space,

Not really. I want to limit the files that wbeserver has to create,
ideally it would only work on a database. Another concern is that with
serve-via-disk scenario on initial http GET request you need to
checkout, serve, clean, while on corresponding POST request you need to
checkout the same revision that was served with GET request, overwrite
with new content that arrives in the request, commit, clean. All
information required to recreate the local copy before commit has to
arrive in POST request. Going through disk seems like a lot of clutter
(also I am not sure if I really know how go about all the details, I
have never used monotone).

> or because you don't want monotone to rescan a whole (arbitrarily large)
> tree every time a user alters one file in it?

It was not my concern. Is this a problem?

BR
Grzegorz