Proposals
Feel free to put blueprint-like or LEP-like Foundations documents here.
Template for Proposals, based on LEP template
Incomplete
45419 ("Launchpad needs a way of easily flagging spam")
Hide and/or Delete Spam: LOSAs need this.
Nominate Spam and Administrate Spam Nominations: We want this. It's a lower priority than the first one.
Invalidating Cached Page Snippets: A possible replacement for the current TALES memcached integration based primarily on invalidation rather than expiration.
Performance Tests in Windmill: Implementing tests to enforce limits on the number of items loaded in a browser by a page initially and on a repeat viewing.
Unique ID for everything: Right now developers can't simply refer to any arbitrary object in the database. This means that library file aliases can't keep track of their parents for security purposes (see 397674), comments can't be implemented generically for all objects, and other issues (collect here, or on page when it is made).
- Remove *.launchpad.net (with redirects for deprecation): Having different virtual servers (bugs.launchpad.net, code.launchpad.net) accomplishes very little, and for a significant cost: we can't share expensive-to-download files (combo.css, launchpad.js, and more) across the boundaries. This means every shared +icing slowdown compounds its cost for each "pillar". What about just using launchpad.net/bugs/ and launchpad.net/code/ and so on, like most other web applications? [Question for mars: why can't we share these again? Can we put these on a CDN? Note: JS on CDN is a security risk.]
- Local "VPN CDN": Robert is going to test XXX
Speed up SQL deployment by making changes more careful/restricted XXX tied together with new version of MergeWorkflowDraft, which is itself tied together with the QA tagging changes
- Switch to using directory of sdists. Make it easy to share eggs across deployments (autogenerate eggs?). Switch deployment to using zc.sourcerelease to package up eggs for minor cleanup of Makefile. XXX
Simplify Merge/Batch-Testing Machinery: implement our batch-testing approach with a single tool (tarmac eventually, maybe PQM as an initial experiment) to eliminate buildbot and poller machinery.
Simplify daily launchpad development by changing to a new MergeWorkflow.