[Gnash-dev] Re: goals for next gnash release

[John Gilmore]

> >                          I think our next release should focus on
> > consolidating the gains we've made, making a bunch of that new code
> > work to give us a working Flash 10 parser.  And testing everything
> > against the top 10 or 20 video sites to knock off the rough edges.
> > What do you think?
> 
> The top video sites rely increasingly on AVM2, so those are more or less
> the same goal. Besides AVM2, getting the core libs to be properly
> reentrant would be an important goal for making Gnash work better as a
> plugin, improving performance, and generally being able to implement
> things more quickly.
> 
> The obstacle to both AVM2 and reentrancy is extending the Relay type
> system to all Gnash's AS objects. This is because the most important
> part of AVM2 initialization is to construct an object, turn it into a
> MovieClip, and use it as the main timeline (root movie). Gnash is
> currently incapable of turning an object into a MovieClip.

Why wasn't this a problem in AVM1?

Is this AVM2 initialization done inside gnash?  Or it it done in an
ActionScript file written by somebody else?

If it's done by us, why don't we just create the object as a MovieClip
from the very start?

If it's done by someone else, is it because there's a stylized way to
start an ActionScript file?  If so, can we make the stylized thing work,
e.g. if they all create a "Foo" object and turn it into a MovieClip,
let's quickly fix our Foo class so it can be turned into a MovieClip.

This would let the internal interpreter restructuring (which would let
ANY class be turned into a MovieClip) happen in parallel with the
other debugging work needed to get AVM2 players working (which Rob or
volunteers can do).

> I started implementing the code necessary to do this, so it already
> works for less complicated objects (and fixed one of the top 20 video
> sites), but until Relays are extended to MovieClip and friends, AVM2
> won't be able to create the tree of DisplayObjects it needs to show
> anything correctly. Doing that really has to be the first priority as
> it's the root of most things that are wrong with Gnash and blocks
> progress on just about every front.
> 
> Once that job's finished it would be easier to work out what's
> achievable for another release. I hope that AVM2 would start to be
> useful for more than tracing almost immediately.

Great!

Is the Relay stuff documented anywhere?  I just looked at it in your
recent checkin, and don't understand why it exists or what it's good
for.  There are no comments.  Can you spend an hour writing something
into the Gnash internals ("reference") manual about what Relay does,
why and how it does it, where it it used today, and where it should be
used?  Why is it even called Relay?

Thanks...

        John