On Aug 21, 2006, at 12:23 PM, Thomas E. Spanjaard wrote:

> I think mycroft's argument is that a concurrent process busy in the  
> same CG dramatically reduces I/O throughput available for te DVD  
> burner. Which is not unfeasible, as in the late nineties CD-R(W)  
> burners had issues with buffer underruns because the host machines  
> couldn't handle the I/O throughput necessary. As DVD 1x is about CD  
> 9x, it's not hard to imagine  this having an influence.

I don't think the concurrent I/O is as much as a problem as the  
interleaving of two files' blocks as stated in his example.  That  
prevents the data from being streamed back to the host effectively  
because the file system is forced to issue a bunch of small reads  
instead of large ones.

-- thorpej