David Farning wrote: >>- It can calculate an optimal mirror for Fedora Core users for the >> updates and base channels from among 130 sites. It doesn't take >> too long either. > > Good start ;) The hurdles that I am facing working on something > similar. I saw a screenshot of your channelSelector linked from a Yum bugzilla entry, looks interesting. The scopes of the programs are still different at the moment IMHO. > What is an optimal mirror? I measure a simple operation of getting some yum headers. Maybe the BBB would sue me for false advertisement, but in my tests this is at least a good try at ranking the mirrors. > Server in same country as client? Some user are billed more for out of > country downloads? The countries are known, but not currently used for any calculations. You can filter the search based on region. > Current number of user on server? Is the server at it's max user limit? Maxed out servers are excluded from the search. > upToDateness(TM)? How many hours is the server behind the primary > server? Good point, the solution to this would probably be quite complicated. > Security? Can we trust the mirror. > I think that these issues need to be looked at carefully before > releasing something like yummy on the world. Most of the mirrors are > mirroring out of generosity to open source movement. We really don't > want to take advantage of them by abusing their mirrors. The mirror list is based on the "official" list of Fedora mirrors from fedora.redhat.com. I think the load definitely needs to shift from download.fedora.redhat.com to the mirrors, I don't see this as abuse. As one mirror gets more users its score is lowered and the tool moves users to the most responsive mirror. Regards, Tomas
