|Re: [egit-dev] retention policy for builds|
It might be worth noting that using composites is not all benefits.
Here are a couple of things that might be worth considering:|
1. Adding a p2.index file to the same directory as the compositeContent.jar (or .xml) will minimize the number of accesses that p2 will need in order to identify the type of repository. p2 always looks for this file first. If it isn't found, it will look for content.jar, content.xml, compositeContent.jar, compositeContent.xml (in that order). For people sitting across the big pool, each 404 in that process costs at least half a second.
2. The performance when downloading meta-data is degraded each time a new repository is added to the composite. The degradation is more or less linear. This can be perceived negatively by your users so if you have a large number of children, it might be better to merge them into one single plain repository.
3. An alternative to the composite approach might be to say that the main (most commonly used) repository is the child that contains the latest stuff and inform the user that a more complete repository (containing all history) can be found using another URL. My guess is that 99% of your users will be happy with your latest release.
- Thomas Hallgren
On 2011-06-15 15:15, Matthias Sohn wrote:
Back to the top