Joerg Sonnenberger wrote:
> On Wed, May 16, 2007 at 12:32:14PM -0400, Greg Troxel wrote:
>> There's clearly going to be some scheme for worker nodes to ask for the
>> next package to build,
> 
> Yes, by asking the master node via TCP. This can be forwarded via ssh or
> other means, if necessary.
> 
>> and to fetch the prereq binary packages that are
>> needed.  I'd like to see this explained, and in particular whether it
>> relies on a shared filesystem.
> 
> The way I am currently using it is via NFS. Other approaches is to just
> rsync the currently available list. Using FTP would be an option as
> well, but has the problem of checking for up-to-date packages being much
> harder. So a shared filesystem is one option, but not required.

In the interests of getting an initial implementation worked out that 
almost everyone could use right away, I think that assuming there is a 
shared file system between the nodes sounds perfectly fine.

	Cheers,

	-- Johnny Lam <jlam@pkgsrc.org>