There are many packages where the sourcecode isn't hosted at the Gentoo servers, but on sourceforge, kde and other servers, some are dead servers others are far away from where you are ang gives a low speed Reproducible: Always Steps to Reproduce: 1. LINGUAS="fi se zh_TW en_GB" emerge kde-base/kde-i18n Actual Results: Download is done from download.au.kde.org which is on the other side of the world. around 8k/s instead of 200-800k/s Expected Results: Use the servers that has been selected with help of mirrorselect All packages that is offered by Gentoo should be hosted on the Gentoo servers, so that users will be able to download things in a resoneble speed.
distfiles is already large, IIRC a complete gentoo mirror (incl. stages and livecds) is around 60 GB. And for packages like kde which provide their own mirror network, I think it is good to not have them on gentoo mirrors, but rather using the mirrors provided by the official package owner. As of portage 2.0.50 you can specify a mirror to be used for specific mirror:// syntax in SRC_URI. If you want to prefer a specific kde mirror you put a line like e.g. kde ftp://download.at.kde.org/pub/kde/ or for sourceforge sourceforge http://heanet.dl.sourceforge.net/sourceforge in /etc/portage/mirrors and portage will first try to download from that mirror and fall back to all mirrors specified in /usr/portage/profiles/thirdpartymirrors See man 5 portage for more information concerning /etc/portage/mirrors
Would int be possible that a tool like mirrorselect to check up which thirdparty servers are fast enough, like it does check the Gentoo servers. But I still think that Gentoo servers should have the complete collection of sources used in Gentoo, otherwise you could save more space by just using ftp.redhat.com and download the SourceRPMs or source packages from any other Linux dists servers. Today harddrive prices are quite low, which makes arguments about space IMHO quite lame.
IMHO the only reason for keeping all sources in distfiles is, that a company might want to have their own local mirror of them. But they could add a local type mirror to /etc/portage/mirrors, download the files by hand and place them on their local mirror. For me one solution is, to extend the features of mirrorselect to not only probe the distfiles mirrors, but to probe all or parts of the mirrors in /usr/portage/profiles/thirdpartymirrors and add the best in descending order to /etc/portage/mirrors (if version of portage is installed which suports /etc/portage/mirrors) Sure, HDD space isn't the issue. But why create duplicate data, if there are enough existing mirrors for the sources. And the sourceforge and kde mirrors aren't that bad.
we may change this with the next release of gentoo. not decided yet. closing as later.