Re: How does gitosis know who the key belongs to

[Howard Miller <howard@e-learndesign.co.uk>]

Hi,

I don't really understand the answers I have been given. Trying
desperately not to "top post" see bellow...

2009/9/24 Matthieu Moy <Matthieu.Moy@grenoble-inp.fr>:
> Heiko Voigt <hvoigt@hvoigt.net> writes:
>
>> On Thu, Sep 24, 2009 at 03:51:44PM +0100, Howard Miller wrote:
>>> Gitosis obviously uses keypairs but the config file addresses the user
>>> by name/host. How does gitosis connect the two together? Is it any
>>> more complicated than the user detail at the end of the public key?
>
> The user@host at the end of the public key is mostly a comment. No
> serious software would use it for authentication, since you can freely
> edit it without changing the rest of the key.
>
>> Not much different. It uses the command feature of the authorized_keys
>> file of ssh to limit access to the 'gitosis-serve' command which is
>> given the username from the gitosis.conf as argument.
>
> ... the username -> key association being done in gitosis's
> configuration itself (the key held in keydir/foobar is the one of user
> foobar).


Yes, I realise that the association is done im gitosis' config file,
but that is exactly my question - how does that association work?

To take a step back - I have to put a name (or a name@hostname) in the
configuration file. How do I know what that should be in the general
sense? When I create a key it doesn't ask me for the username I want
so I see a disconnect here. Do I put a name, do I put name@hostname,
does it matter and (again) how does it know which is 'my' public key?
Is it the naming convention of the key or something magic? Plus, when
I actually use gitosis, I am logging as the user 'git' so how does it
know who I am? I am thoroughly confused by the whole thing. I'm
probably missing something really obvious but, AFAIK, this is not
covered at all in the docs.

Thanks!

Howard