On 02/03/2010 03:19 AM, Marius Mårnes Mathiesen wrote:
>     But you guys have your own git-daemon that is run. Such access control
>     could be coded into this, correct?
> 
> 
> Actually, to my knowledge, the git protocol itself lacks authentication
> support; it is designed for fast read access to repositories. 

I'm not talking about the git protocol; I'm talking about the git-daemon
process itself. Since you have a custom git-daemon, it's conceivable
that part of that customization could be involve access control
controlling whether the daemon actually responds to a particular client.

> Newer versions of git actually have much improved HTTP support; it is
> faster and supports writing (ie push). So HTTP is a real alternative to
> SSH these days, as long as the users have recent Git clients.

HTTP has supported push for a long time, at least a year. But it doesn't
(or maybe didn't) work very well. Client setup could be a pain, and
worse, the bare repos on the server often wouldn't update properly,
forcing manual intervention to do things like have them garbage collect
(I once saw such a bare repo being updated over HTTP balloon from 30MB
to 2GB).

When I brought these issues up in #git I was told that HTTP was
basically a second class citizen and that nobody really was caring much
about it.

I'd advice much testing, and caution...

--Jeff

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to