So does raspbian (though the urls we post on the cenus page are the "main server only".I note that surprisingly Kali seems to use mirrorbrain.
IIRC in raspbian we are currently using manually built packages (there were some library dependency issues iirc) built from dscs supplied by mirrorbrain upstream. We are currently running a wheezy system and I will need to investigate the implications on mirrorbrain of upgrading to jessie in the not too distant future.Mati, has Kali packaged mirrorbrain or are you using a manual setup?
The main thing with mirrorbrain is you can't safely use it on the dists directory because of it's lack of mechanisms for correctly handling files that change in place. We handle this simply by serving the dists directory off the main server without any redirection obviously this comes at a price in bandwidth from the main server..Have you found any problems with using it for serving apt repositories?
Back when we were looking to set up a redirection system for raspbian mirrors I discussed the http.debian.net code with Raphael Geissert. After some discussion I concluded that it was designed a large network (e.g. multiple mirrors in each country so the behaviour of removing a mirror from rotation during updates didn't cause too much pain) of well-run mirrors (e.g. updated regularlly, updated using the debian ftpsync script which provides metadata that the http.debian.net code needed). Afaict it works on a basis of redirecting all a users traffic to a given mirror which means outdated mirrors are useless.Some of you may have seen the announcement about the Debian http redirector, which is a special redirector for apt repositories that was developed to suite the needs of Debian mirroring where there are many mirrors of varying updatedness, location and quality.
We had a smaller network of mirrors (so losing a countries only mirror because it was updating didn't appeal) and many of them were using plain rsync (which afaict doesn't work with the http.debian.net redirector code) and/or weren't nessacerally updated regularlly. So we decided mirrorbrain with it's file-based approach was a better fit.