On 20/04/2012 Mark Ramm wrote:
* Route "autoupdater" traffic through one system (MirrorBrain)
* Route web based traffic through another (SF as primary, and Apache
mirrors as secondary)
Note that "autoupdater" traffic is probably still web-based traffic: I
cannot check now, since updateXY.services.openoffice.org still has to be
restored, but one of the use cases -not sure it is the only one- is that
the dialog just takes you to a web page where you can select your
favorite version and download it.
So we would need two separate "endpoints" on the website, corresponding
to the two different systems:
- /update/... for MirrorBrain
- /download/... for SF/Apache
Since the downloaded files will be exactly the same (OpenOffice does not
support incremental updates, you must re-download the whole package)
there's also the risk that mirrors are asked to carry two copies of the
same file.
But, in the end, I would prefer this solution, that manages to have
clear separation and keeping systems alive, to the idea that MirrorBrain
is "legacy" technology not to be used for distributing OpenOffice 3.4.
MirrorBrain is excellent, reliable, community-driven technology and it
deserves a place here. If Apache can embrace it, even better: there have
been very promising exchanges between Peter and Infra on this list.
Regards,
Andrea.
