Community
Participate
Working Groups
When searching for features, nested features are also downloaded. Update should not download the feature jar unless the feature has been selected for install.
This has slightly been improved by not downloading the nested features of features that will be filtered out for other reasons.
I've tried to think about it, and I believe that the solution would be to put more information about the available features in the "site.xml", practically duplicating a lot of info that's available in the feature.xml for all the features. Let's compare with the way blogs/webssites works with an RSS feed. I think we can compare the "site.xml" to the "rss.rdf": they are both a summary of what's available on a site; you can use it directly, by displaying the content, and/or drill-down to get more info (the complete article for rss, the complete plugins for site.xml); and the client can compute history, see what he has already downloaded, to show only the new stuff to the user. RSS files duplicate a *lot* of information that's available on the html pages, and that's perfectly OK because the goal is to make it easier for clients to access a summary, an index, of the site. I believe it's exactly the same for site.xml: it's a summary, an index, of what's available. And an index's role is also to ensure acceptable performance. Since site.xml is completely under Eclipse.org's control, I think it would be OK to add all the info to it. Then, looking at an update site would become a snap. (I'm even being told that there's an extension point defined to extend site.xml - I haven't looked myself though.
i agree. site.xml should be extended or some other file introduced that ha more information about features so that feature jars do not have to be downloaded.
*** This bug has been marked as a duplicate of 144876 ***
