[RFC] permissions mgmt

Martin Pool mbp at canonical.com
Wed Aug 23 05:21:52 BST 2006

On 21 Aug 2006, Robert Collins <robertc at robertcollins.net> wrote:
> I think the bunch of interactions between permissions and performance
> are a hint :).. and our inability to do the right thing on sftp [at all]
> hints this too...

I can see that hint too.

> I'd like to propose the following changes to our permission support
> logic:
>  - over sftp we dont chmod at all - we do whatever the umask is
> configured to do by the user. We should offer a helper script to set the
> umask before invoking the sftp-server process.
>  - over ssh we offer a umask-setting facility. We will have a native ssh
> non-smart server protocol in 0.11 I think - its the substrate for the
> smart-server-over-ssh.
>  - on local disk we offer a umask setting facility.
> Both the umask-setting facilities are off by default, controlled by a
> file in the repository.

So, if the file exists, the process will just do umask(x)?  It sounds
reasonable to me.  It has the slight sideeffect that it will also
affect any working trees which are created, etc, but it has the benefit
of simplicity.  I suppose you could do it for specific sections of the
program which need to interact with the repository.


More information about the bazaar mailing list