[MPlayer-dev-eng] Fixes/enhancements to vdpau

[Jason Tackaberry]

On Thu, 2009-04-09 at 19:47 +0300, Uoti Urpala wrote:
> VO-specific functionality to allow fast overlay updates. You insisted on
> a completely VO-independent implementation, an approach which I
> considered unreasonable as you were willing to accept more important
> drawbacks such as requiring the use of -nodouble.

Although I readily admitted that the requirement of -nodouble was broken
and temporary, and had said that I would work to address that once I
submitted the overlay patch for comment.

The discussion died after the consensus of the core developers (notably
you, Michael, and Rich) was that the VF should just be responsible for
overlaying per frame, and any guarantees of latency be implemented as
VO-specific extensions.

I did not reject this suggestion.  I merely went away to regroup.

I still nevertheless think it's accurate to conclude from the experience
and other threads on mplayer-dev-eng that any patch that touches core
parts is much more likely to undergo heavy scrutiny than one that
strictly touches a vf.  And that seems sensible and intuitive to me.


> Now you were complaining about the previous discussion while talking

Characterizing my comment that "MPlayer hackers do not like it when
patches touch core areas like timing" as complaining is obviously a
matter of perspective.  I can see why you, as an MPlayer hacker, might
interpret it that way.


> about your new implementation that not only requires VO-specific
> support but in fact only works with one custom VO.

If all other things were equal, it certainly would be inconsistent of me
to have an implementation that works with one custom VO.

But basically nothing is equal, and the new approach is completely
different from before.  I have come to realize that the philosophy of
using MPlayer as what is effectively a display manager is fundamentally
flawed.

Regards,
Jason.