On 18.06.2016 12:39, Ned Slider wrote: > > > On 17/06/16 20:04, Warren Young wrote: >> In another recent thread,[1] someone was having trouble with the >> yum-plugin-fastestmirror feature, so I suggested he remove it, since >> it’s just a plugin and should therefore be optional. He reported >> that it couldn’t be removed due to package dependencies. >> >> I investigated further and found that this also affects CentOS 5.11 >> and CentOS 7.2. (The OP is on CentOS 6.8.) So, I reported it as a >> bug upstream, and they claim their yum package doesn’t do that, so >> they’re bouncing the problem back downstream. >> >> I checked, and they’re right insofar as CentOS’s current C7 yum.spec >> file has this on line 118: >> >> Requires: yum-plugin-fastestmirror >> >> I then tried to go get the RHEL SRPM for yum to compare its spec file >> to C7’s, but their download site just refers visitors to >> git.centos.org, and the yum repo there doesn’t seem to have an >> upstream RHEL7 branch. >> >> So, I started poking around in the yum.spec file history, and indeed, >> the oldest yum.spec file on the c7 branch doesn’t have that Requires >> line! It was introduced in checkin f1c1b982, which claims all it >> does is “debrand yum-3.4.3-132.el7”.[2] >> >> I realize it is in the CentOS project’s best interest if users always >> use the fastest mirror when downloading, but I claim that it is a bug >> to mark any plugin as Requires, particularly when upstream does not. >> > > Correct, the fastestmirror plugin is a requirement on CentOS (and > Scientific Linux) but is not on RHEL. > > It was originally discussed here: > > https://lists.centos.org/pipermail/centos-devel/2008-June/002864.html > > Personally, I'm firmly in favour of having fastestmirror plugin as a > dependency in yum as it will benefit the vast majority of people IMHO. > Plus it is easily disabled for those who disagree. > yes but why is it not possible to have really the fastest mirrors? I'm from Europe, and with or without this plugin there are only hosts from other continents are contacted ... and that there would be no host in Europe is definitely wrong ... as it seems, the configureable parameters in /etc/yum/pluginconf.d/fastestmirror.conf have never been tested before ... because exclude=.edu,.com,.org include_only=\.at$,\.ch$,\.de$,\.nl$,\.uk$ gives this: Determining fastest mirrors * base: mirror.keystealth.org * epel: mirror.math.princeton.edu * extras: mirror.keystealth.org * updates: mirror.keystealth.org why not mirror.nextlayer.at or mirror.inode.at? so the question where can I really take influence which mirrors are used? or just how is "fastest" defined? Thanks, Walter