Re: [Vnc2swf-users] Project Future

[Yusuke S. <yu...@cs...>]

Hi,

To create a speed-efficient recorder, we really need faster byte operations.
Probably the most expensive part in the current pyvnc2swf implementation is
just to convert the whole bitmap into another format. 

More specifically, look at line 61 in image.py:

def create_image_from_string_argb(w, h, data):
  data = ''.join([ data[i+1]+data[i+2]+data[i+3]+data[i] for i in xrange(0, len(data), 4) ])
  return pygame.image.fromstring(data, (w, h), 'RGBA')

Now imagine a VNC server sends you a 1600x1200 bitmap (7.6MBytes)
That's a lot of data for doing this kind of operation for each frame.

For some time I was thinking of writing some external functions in C, which
would be hundred times faster to do this kinda stuff. But it breaks
the beauty of platform-independentness.  This is the point where I was stuck.
People would need to install a C compiler to build pyvnc2swf. 
Perhaps it's not a big problem in Linux or *BSD, but I was afraid it brings
some extra trouble. I certainly expect complaints from Windows/Mac folks.

So... how do you guys think? Maybe I should ask people what kind
of platform they are using. If cross-platform is not a concern, I
think this is the way to go.

Yusuke


On Thu, 16 Apr 2009 22:33:16 +0200 (CEST), Olivier Hervieu <oli...@wa...> wrote:
> Great, 
> 
> TheAmigo, i think that our needs are the same ... we have to record several (more than 2!!) vnc session on the same computer... 
> 
> your idea to write a ffmeg plugin is great that might interressed people who wants to have an authoring/transcoding tool shipped with pyvnc2swf. 
> 
> So for the next week i propose that everybody interressed in the project of a safe and fast vncrecorder to send on this mailing list what should and shouldn't be the the vncrecorder. 
> 
> I propose the same for people who are interressed in an authoring tool to send here what should be this vncauthoringtool. 
>