svn commit: r306755 - head/Mk

[Emanuel Haupt]

"b. f." <bf1783 at googlemail.com> wrote:
> On 10/31/12, Emanuel Haupt <ehaupt at freebsd.org> wrote:
> > "b. f." <bf1783 at googlemail.com> wrote:
> >> On 10/31/12, Emanuel Haupt <ehaupt at freebsd.org> wrote:
> >> > Author: ehaupt
> >> > Date: Wed Oct 31 18:11:33 2012
> >> > New Revision: 306755
> >> > URL: http://svn.freebsd.org/changeset/ports/306755
> >> >
> >> > Log:
> >> >   Update MASTER_SITE_SOURCEFORGE
> >> >
> >> >   Feature safe:	yes
> >> >
> >> > Modified:
> >> >   head/Mk/bsd.sites.mk
> >> >
> >> > Modified: head/Mk/bsd.sites.mk
> >> > ==============================================================================
> >> > --- head/Mk/bsd.sites.mk	Wed Oct 31 18:02:55 2012
> >> > (r306754) +++ head/Mk/bsd.sites.mk	Wed Oct 31 18:11:33
> >> > 2012	(r306755) @@ -1125,11 +1125,10 @@
> >> > MASTER_SITE_SAVANNAH += \ .endif
> >> >
> >> >  # List:
> >> >  # http://sourceforge.net/apps/trac/sourceforge/wiki/Mirrors
> >> > -# Updated:	2012-10-26
> >> > +# Updated:	2012-10-31
> >> >  .if !defined(IGNORE_MASTER_SITE_SOURCEFORGE)
> >> > -.for mirror in heanet sunet iweb switch freefr garr aarnet \
> >> > -		voxel jaist osdn nchc ncu internode waix \
> >> > -		hivelocity superb-dca3
> >> > +.for mirror in aarnet freefr garr heanet ignum hivelocity
> >> > internode iweb jaist \
> >> > +	nchc ncu netcologne osdn space sunet superb-dca3 switch
> >> > tenet ufpr waix
> >>
> >> Dmitry and Pav went to some trouble to sort this list so that those
> >> mirrors resulting in faster downloads for a majority of users were
> >> placed first.  Have you verified that an alphabetical sort won't
> >> result in slower average downloads for many users?
> >
> > I was not aware of that. I just noticed that the list was somewhat
> > outdated.
> >
> > A site that's fast for you might not be fast for someone else due to
> > the geographical location so I am not sure what criteria was used to
> > determine which one was better than others.
> >
> > One can always use MASTER_SORT to priorize some sites.
> 
> Yes, of course -- but here we are concerned with the average speeds
> using default settings, for the largest number of users.  We can't
> provide a fixed ordering that works well for everyone, but the
> heuristics Dmitry used last time are mentioned throughout the thread
> starting with:
> 
> http://lists.freebsd.org/pipermail/freebsd-ports/2009-September/056783.html

Interesting stats. Thanks for finding that post.

I've restored the previous sort order for now.

Emanuel


> One thing I noticed in my last change to the Sourceforge list is that
> some mirrors are still performing well despite being taken off
> Sourceforge's list of official mirrors.
> 
> b.
>