www-mirrors mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew Kenna" <andr...@stamina.com.au>
Subject RE: cvs commit: site/xdocs/dev mirrors.xml
Date Wed, 27 Nov 2002 23:25:11 GMT
Thanks for the other ideas, I was trying to get people thinking of what
we could use to categorse quality mirrors...

With the point of people relying on binaries, I'm reffering to people
that use up2date from redhat and assume that it will update their apache
daemon.. It might but it only tags the version as 1.3.22 for instance..
Or one other case I heard about in there was debian patching up 1.3.9..

I've had this discussion with Joshua before, but I think if people are
serious about having a quality mirror they should download the source
code from apache.org or an apache mirror.. Compile it up and be done
with it.


-----Original Message-----
From: jason andrade [mailto:jason@dstc.edu.au] 
Sent: Thursday, 28 November 2002 10:17 AM
To: mirrors@apache.org
Cc: infrastructure@apache.org
Subject: RE: cvs commit: site/xdocs/dev mirrors.xml

On Thu, 28 Nov 2002, Andrew Kenna wrote:

> I guess another question I'll ask, how are we going to define high 
> quality mirrors ?
> A) Based on bandwidth

it would certainly be one of the factors.

i think the primary ones would be

o bandwidth

o ability to carry a complete mirror - e.g about 15G of disk space
  to cover every apache project ?  plus the daily updates that entails.

o ability to refresh more than daily - perhaps to implement a ssh
  based signalling system similar to that in use by gnome and other
  projects, to pull updates when they are in place on the master site

o ability to act as a propagation server - e.g running rsync for
  other downstream mirrors

o responsiveness of admin contact required to deal with mirror

> B) whether the person relies on pre-packaged binaries to update their 
> web site or goes out downloads the source and compiles it themselves ?

i personally disagree strongly about individual "mirrors" doing this
because then they are no longer mirrors.  if people are providing
binaries they should be submitting them to the official site where it
then propagates out to all the other mirrors.

> C) Restrict the number of mirrors in each country to say 4

4 might be a bit small, but i guess there might be some policy there.
in the US it might make sense to have 15 mirrors.  in australia 6.  in
Tibet, 2.  it would be a (clearly understood) function of network and
demand (number of internet users) in that region.



View raw message