Am 10/26/2011 11:57 PM, schrieb Peter Pöml:
Hi,
On Wed, Oct 26, 2011 at 01:43:20 -0700, Dave Fisher wrote:
Are there requirements for more than what is on the mirrorbrain page?
The page should list all essential requirements.
http://mirrorbrain.org/requirements/
Extracting the list gives:
Apache HTTPD server
mod_mirrorbrain
mod_form
mod_geoip / libGeoIP
PostgreSQL (or Apache DBD API compatible driver)
While the DBD framework would allow to use other databases, MirrorBrain
depends on features that only PostgreSQL offers. So this is a must, I'm
sorry ;-)
mod_asn ?
Highly useful when there are many mirrors, and when downloads are large.
Both is the case with OOo. Thus, we have used mod_asn for OOo downloads
so far and I would really recommend to continue using it.
If we are on ASF hardware then we won't want to serve a local copy:
So, "Alternatively, a pseudo file tree can created locally. In that case,
MirrorBrain needs to be configured to never deliver files directly."
http://mirrorbrain.org/archive/mirrorbrain/0045.html
MirrorBrain is more than a file server or redirector. It automates
the creation of checksums, hashes, torrents, metalinks and serves
cryptographic signatures. Files need to be present locally to exploit
these features. These features are useful for OOo because downloads are
generally large, which requires reliable transfer.
@List:
Has anybody an idea about where to host this service? It doesn't need to
be necessarily inside the ASF.
Really nobody?
If the mirrorbrain instructions are that simple and can be easily supplemented
with detail.
Setting up MirrorBrain would be one way, but it would require
replication additional configuration (for instance, download statistics)
that we have built on the current download server
(download.services.openoffice.org).
Another way would be to simply have a virtual machine, where we move the
current server to. That would cause the least effort, I guess.
+1 this should be really our goal for now ...
Starting from scratch would mean to lose a lot of the previos work --
and I really mean lots, which I dare to judge because I spent a lot of
time with download.services.openoffice.org.
On the other hand, having MirrorBrain at the core of the ASF's mirror
system could be interesting for other projects, too. I know closer.cgi
but I'm sure that MirrorBrain could serve the ASF well. (Well possible
as an addition, rather than a replacement, for a soft transition.)
That might outweigh the pain of creating OOo's download service from
scratch in a different environment.
... and this the long term goal.
The ASF can really benefit from this way of downloading software.
Currently you have to choose a mirror, then change to the directory
structure where the respective binary is located and finally download it.
OK, the target audience the ASF currently has is developer-oriented, so
this is not too difficult for them. ;-) However, it could be indeed easier.
But I try to think more as an enduser. And these people need to have a
simple solution. That means a link on a webpage where they click on and
the download of the file is starting. IMHO we have already a very simple
solution you can see on "http://download.openoffice.org"; and I would be
happy to help to buildup a similar thing for the entire ASF.
@Donald:
Of course we can keep the current download behavior to satisfy also
users that needs to download from a specific mirror server.
Marcus
Are you involved in maintaining ASF infrastructure, or know people who
are?
Like what database is currently being used on download.s.oo.o? and which
version?
Uhm, PostgreSQL 8.4.4. (But any 8.3/8.4 version should do.)
Thanks,
Peter
