Re: support for private repo on AWS S3

On Mon, 28 Sep 2009, James Antill wrote:

> Jeff Enderwick <[email protected]> writes:
>
>> This is what I meant by AWS auth:
>> http://docs.amazonwebservices.com/AmazonS3/latest/index.html?RESTAuthentication.html
>
> Ok, looking around I see a patch from 2007 for libcurl support:
>
> http://curl.haxx.se/mail/lib-2007-04/0239.html
>
> ...and if that got in then newer urlgrabber having an option to use it
> shouldn't be too hard. But it isn't in the libcurl in Fedora atm.
> For "direct" support, if someone really wanted, they could probably do
> what the RHN plugin does ... but that's not going to happen from
> upstream.
>
>> Does S3 support HTTP Basic authentication? I can't seem to find any refs to
>> that. Are you assuming a web server between the yum client and S3? I'm
>> looking to have the EC2 instance directly pull from a repo stored on S3.
>> Maybe my google skills are broken :-)
>
> I don't really know anything about how special EC2/S3 is. I know
> people have working setups where they have an SSL box on the server,
> and use SSL+Auth. Basic on the client. I also know a couple of people
> using x509 certs on their clients, with checking in their servers (on
> RHEL-5.x).
> So I'd look at either of those options, if you want to solve it in a
> easy/std. way.
>
>> I'm currently using the rightscale centos 5.2 ami.
>
> 5.2 yum is pretty old, I think you need the 5.4 yum for
> x509. Although using libcurl will be a RHEL-6 feature.
>

But if you're  adventurous and don't mind updating pkgs from stuff in 
fedora, you could try out urlgrabber 3.9.1. It was intended as api-compat 
with earlier urlgrabber so we wouldn't have to change any of the callers.

-sv