Return-Path: Delivered-To: apmail-mirrors-archive@apache.org Received: (qmail 55138 invoked by uid 500); 26 Apr 2003 15:43:18 -0000 Mailing-List: contact mirrors-help@apache.org; run by ezmlm Precedence: bulk Reply-To: mirrors@apache.org list-help: list-unsubscribe: list-post: Delivered-To: mailing list mirrors@apache.org Received: (qmail 55126 invoked from network); 26 Apr 2003 15:43:17 -0000 X-Epoch: 1051371798 X-Sasl-enc: qDyc0oVcgtbcAR2f6sOpJw Date: Sat, 26 Apr 2003 11:42:55 -0400 (=?ISO-8859-1?Q?Est_=28heure_d'=E9t=E9=29?=) From: Joshua Slive To: mirrors@apache.org, Ads Administrator Subject: Re: Default mirror preference? In-Reply-To: <012901c30c09$417bb090$cb00a8c0@SYSTEM8> Message-ID: References: <20030426035831.GA23770@timmy> <012901c30c09$417bb090$cb00a8c0@SYSTEM8> X-X-Sender: slive@www.fastmail.fm MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII X-Spam-Rating: daedalus.apache.org 1.6.2 0/1000/N On Sat, 26 Apr 2003, Ads Administrator wrote: > Hi, > > We are one of the official mirrors in Hong Kong. We were wondering if > there are default "preferences" given to particular mirror sites? For > example, if a particular site has higher bandwidth than another, is > perference given to the higher bandwidth site (eg. the higher bandwidth > mirror is selected as the default more times than the lower bandwidth > mirror sites)? At the moment, the preference system is binary. It works like this: If the mirror was reachable during our last test and the mirror has been updated in the last 48 hours, then the mirror is available for preferred status, otherwise it isn't. > > If this isn't done... would it be a good idea? This information could be > based on the information the mirrors give when they sign up (or could be > submitted separately if required). What do you think? Certainly it is a possibility. The issues I see are as follows: 1. Is there a problem here to solve? ie, are people getting badly performing mirrors because of the current system? If so, should we just remove the poor mirrors rather than use a more complex system? Unfortunately, I don't see any way to gather statistics to answer this question. 2. We don't have very good communication with a large chunk of our mirror sites. Not everyone is on this mailing list, so it might be difficult to get data on mirror capacities. We are always, of course, open to suggestions and further discussion. Joshua.