I think this is a good idea. Start with /home/admin or /home/a/admin
for that matter and extend it later.

the run-crons script does not switch users, the user running the
scripts has been defined in the /etc/crontab file. In some way this
resembles our setup:

cron-daemon checks system crontab and user crontabs and executes jobs
in system crontab with configured user and in user crontab with the
specific user rights. The /etc/cron.d directories are a helper,
executed by another service that runs with configurable permissions.

regards,

Lars

On Fri, Dec 19, 2008 at 4:56 PM, Bertrand Delacretaz
<[email protected]> wrote:
> On Fri, Dec 19, 2008 at 4:38 PM, Lars Trieloff <[email protected]> wrote:
>> I doubt that. On my Gentoo box with Vixie-Cron I can find following
>> statements in the crontab:
>>
>> */10  *  * * *  root    test -x /usr/sbin/run-crons && /usr/sbin/run-crons
>>
>> which means it is more or less hardcoded that root runs the script,
>> and relys on the correct system setup....
>
> I don't know this run-crons script but if it runs scripts for users
> other than root it might as well use the script's user owner identity.
>
> Anyway, my observation applies to the general case of running a script
> outside of the context of a request.
>
> As others have suggested, requiring that such scripts be placed inside
> a user's home folder, and assuming that that folder is properly
> secured, is probably the easiest way to define the identity under
> which to run the script.
>
> An additional safety feature might be to only accept scripts that have
> specific access rights, for example "only writable by owner".
>
> If we want to use non-request scripts only for admin initially, we
> might require them to be stored under /home/admin, so that we can
> later allow other users as well.
>
> -Bertrand
>
>> On Fri, Dec 19, 2008 at 2:31 PM, Alexander Klimetschek <[email protected]> 
>> wrote:
>>> Not exactly. To cite Bertrand:
>>>
>>> Unixish systems solve this by using the identity of the user who owns
>>> the script (unless the setuid flag is set), and enforcing the way this
>>> identity can be set - but we don't have that kind of feature in JCR,
>>> or do we?
>

Reply via email to