On Wed, Jun 06, 2007 at 05:11:54PM +0200, Duncan Mac-Vicar Prett wrote:
On Wednesday 06 June 2007 16:44:34 Michael E Brown wrote:
Mirrorlist functionality is a really useful capability. I dont see how you could you could support millions of users without it. It allows you to load balance between separate servers which may even have different paths to the repo. It also enables you to do really cool things, like have a cgi script that produces the mirror list. It can, for example, return geographically-close mirror based on the users client IP.
Variable sustitution in urls looks really interesting. I will try to implement that.
mirrorlist, why not handling it at the web server level?
Several reasons. - In the current yum implementation, the client gets the mirror list and randomly chooses an entry from the list. - It also will fail over down the list if there is any connection problem with the first server. - The mirror list can list different servers that have the repo at their own path. - Some people dont have administrative access to the web server to change the web server configuration to redirect. The mirror list is simply a URL to a file containing a list of mirror paths. Because it is a URL, it can easily be configured to point at a cgi that outputs the list. -- Michael -- To unsubscribe, e-mail: zypp-devel+unsubscribe@opensuse.org For additional commands, e-mail: zypp-devel+help@opensuse.org