| [cross-project-issues-dev] p2 metadata on your update sites |
Attention cross-project release engineers: Although p2 contains backwards compatibility code for reading site.xml and feature.xml files from classic update sites, there are performance benefits for clients if you "p2-ize" your update sites.
Problem:
- When the user connects to an UM update site that does not contain a site digest, all the feature JARs will be downloaded which will have a "less than optimal" consequence in term of user-experience and performance.
Solution:
- Generate p2 metadata for your update site:
- For static update sites (e.g. sites in which the content is no longer changing, like 3.3 update sites), simply run the p2 metadata generator as outline on the wiki here: http://wiki.eclipse.org/Equinox_p2_Metadata_Generator#Generating_metadata_from_an_update_manager_site
- For sites which are regularly updated, you can either use the same generator after you have produced your update site, or you can use PDE-Build's capabilities to generate p2 metadata as part of your build process. See Help > PDE Guilde > PDE Advanced Topics > PDE Build Integration with p2 for more details on build integration.
Note that generating p2 metadata is a fairly straightforward operation and simply produces two additional repository index files which live on the server along with your site. The Platform and Modeling teams have already had success in integrating metadata generation into their build process.
Benefits:
- The files resulting from the generation will be fairly small and can be downloaded quickly
- At install time the p2 resolver will validate the inter bundle dependencies to ensure that the system being installed is runnable.
PaScaL