At this point we believe that there could be several causes sometimes inter-related to failures of updating:
Handling of the HTTP error code 403, which got addressed in 3.5.1.
NTLM proxies, fixed in 3.5.1
Order in which the repository factories are being consulted
Foundation redirecting the obtention of repository indexes (content.jar, artifacts.jar, composite...) to servers on the cloud but where the files would be missing from.
The one scenario Ian seems to have notice involves p2 trying to contact the foundation for a site.xml, being redirected to "the cloud" and have the server being contacted return a error code 403 rather than a 404 since the file is not present.
What is planned on the installer?
Henrik briefly chatted about what is available in Buckminster.
Technology to talk back to the server and obtain content for the wizard. Pascal wonders if this could be done with meta requirements or if those would happen too late.
Ability to address several profiles from one request. We will likely need to talk again about such a mechanism in p2 as similar goals are being discussed in the context of OSGi RFP 121 (Applicaiton Lifecycle and management).