wiki:Fellowship2013/Priorities

Version 1 (modified by exarkun, 17 months ago) (diff)

summarize twisted bof discussion

  1. Tickets already in review before
    • Many tickets have had some work done on them but have not yet been completed. These tickets represent a large amount of invested effort that has yet not yielded any benefits.
    • As such, completing them may result in proportionally more return for effort.
    • Some judgement needs to be applied in selecting tickets from this category. Just because effort has been invested does not necessarily mean the tickets are particularly valuable to resolve.
    • Some tickets are stalled or abandoned because they are really, really hard (and so probably require lots and lots more effort)
    • A trac report will produce a list of tickets in this category - https://twistedmatrix.com/trac/report/16
  2. Deferred cancellation
    • Ultimately the vast majority of Twisted-provided APIs that return a Deferred should implement cancellation logic rather than relying on the fallback behavior provided by Deferred itself.
    • This is a very large number of mostly small, self-contained tasks.
    • Potentially a GSoC project or OPW project instead of a Fellowship project (lends itself well to beginners)
    • Most tickets for this work have not yet been created
  3. Deploy to new tummy machine (ASAP)
    • Cube (hosting most current infrastructure) will be desupported in the immediate future.
    • New-Cube is up, has more modern OS, faster hardware.
    • Services need to be migrated. Incomplete list:
      • frontend website
      • buildmaster
      • highscores
      • mail system
      • mailman
      • trac
      • user accounts
      • dns
  4. Attempt trac upgrade in a way that allows switching back to current version
    • Basically, set up a testing deployment/environment where trac changes can be tested before going live
  5. Support for non-OpenSSL SSL libraries
    • API replacement for twisted.internet.ssl Certificate APIs, and transport layer implementation
    • Alternative backends include win32 APIs, OS X SecureTransport (aka libsecurityssl?) and the cross-platform libnss3 (Mozilla).
    • A good starting point would be a API specification, that at minimum includes current functionality.
  6. twist: command-line without daemonization
  7. import krondo tutorial into twisted (if davep agrees), update to newer code
  8. lore/sphinx migration work
    • explore a lore output plugin that generates sphinx output via docutils parse tree objects - use docutils and sphinx as libraries
      • avoids single cut-over event
    • explore a lore output plugin that generates ReST via docutils parse tree objects
      • still involves single cut-over event, but maybe produces better output than previous attempts