blob: 25190f8ee010baf7867424865b71a27c09c248f8 [file] [log] [blame]
short term core API stability
Changes that probably impact the API, carefull discussion (IRC) + design doc is required
before changes are accepted.
target release ! description
0.4.1 ! expose and API to query the supported seek formats/flags on
(done) ! pads, somthing like an extra arg to gst_pad_set_convert_function
! and gst_pad_set_event_function with some function to query the
! flags and formats. more ideas in docs/random/wtay/query_events
! (API: medium dificulty)
0.4.1 ! add event for segment playback/looping and seeking (docs/random/wtay/segments)
(done) ! (API: medium dificulty, plugins: HARD to very HARD)
? ! add event to adjust rate (reverse playback, slow motion, frame skipping)
! (docs/random/wtay/rate_event)
! (API: medium dificulty, plugins: HARD to very HARD)
? ! add method in the scheduler to set the entry point (frame stepping?)
! (docs/random/wtay/scheduler_entry)
! (API: moderatly EASY, scheduler implementation MEDIUM)
0.6.0 ! create a design doc for a timecache implementation,
(done) ! (docs/wtay/random/timecache)
! (API: MEDIUM, needs lots of discussion, plugin implementation MEDIUM to HARD)
! (done: implemented using GstIndex base class + subclasses)
? ! implement a QoS event and a policy for handling the event.
! (API: kindof EASY, plugins MEDIUM to very HARD)
0.4.1 ! implement user defined GstFormat values, make a format factory etc..
(done) ! (API: MEDIUM, plugins MEDIUM)
? ! strip the API to a bare bones minimal set of functions, leave the automatic
! stuff to the app instead of forcing a policy in the core.
! create a library with useful higher level function for people who don't want
! to deal with the lowlevel stuff.
! (HARD, need to negotiate with people :))
? ! use GMarkup to load/save objects, remove dependency on libxml2
! (MEDIUM) breaks API/ABI compatibility
shortterm core feature enhancements
0.4.1 ! Implement PAD_DISABLED. This requires simple checks in the scheduler so that
! it doesn't try to pull/push data from/to a disabled pad.
! When an element goes to the PAUSED state, all of its pads should be disabled.
! This should also work for ghostpads.
! (API: MEDIUM to moderatly HARD, requires some scheduler understanding)
short term usability
Writing docs is NOT boring, you learn a lot, you get insight in stuff, you help a lot
of people, hey! you might even find YOUR book on a shelf in a bookstore!!
? ! plugin writers guide
! (we have almost nothing, so any start is welcomed)
? ! app writers guide needs to cover common tips and tricks and HOWTOs
short to midterm policy definition
Policy definition is closely related to a HOWTO but sometimes needs some thinking.
? ! document thread safety guidelines, what stuff needs locking in the app, what
! is done in the core.
! most of this stuff is in the heads of various people but needs to be written
! down so that people get more insights into the design and vision of GStreamer.
! (MEDIUM, some research and discussion needed)
? ! a step by step guide to the implementation of various events in a plugin, what can you
! do, when is data passing forbidden etc..
! (MEDIUM, some research needed)
? ! figure out a policy for the NEW_MEDIA event
? ! figure out how to better handle clock resync
midterm feature enhancement
0.6.0 | Define and implement a syntax in gst_parse to handle filtered pad connections.
(done) | (MEDIUM)
? | Figure out a way to set properties on schedulers (and bins?) from gst_parse.
? | Make gst-inspect do inspection of plugins, schedulers, autopluggers, types.
| An idea would be to make -inspect output an XML representation of the objects
| and use XSLT to transform this into text, HTML, docbook, ...
midterm (longterm) optimisations
These optimisations can be done without changing the existing API.
(in progress) ! implement an optimal scheduler that uses chaining when possible
! (HARD, requires detailed knowledge of element scheduling)
? ! alternatively optimisations to the current scheduler can be done such
! as: do nothing when the pipeline structure (or chain) has not changed
? ! GstQueue is a little mess. implement a better queue (lockfree?), refactor
! queueing policy (max buffer, max time, levels etc..)
! (MEDIUM to farily EASY)
longterm feature enhancements
Various features that are not critical yet.
? ! factory aliases. map a generic name like "videosink" to and actual
! user configurable plugin (aasink, sdlsink, xvideosink, ...)
? ! property proxy in compount elements. not sure if it's possible at all.
! what with elements with the same property?
! (MEDIUM, needs some thinking)
? ! Make _pad_select work for muxers
needs consensus
Some stuff that needs to be figured out based on a pro/con comparison.
? ! can we decide on the fact that downstream events are traveling using the
! scheduler? do we need to reevaluate that design decision?
! (MEDIUM, needs pros vs cons document)
Benchmarks are always good to get acceptance in a wider comunity or to identify performance
problems that need fixing.
? ! do a latency comparison with popular other frameworks, document GStreamer
! overhead.
! (MEDIUM to somewhat EASY)