RE: IP over DVB-S2

[Jérôme Lacan <jerome.lacan@ensica.fr>]

> Due to the different length sizes for the data fields, 
> clearly the first objective is:
> - Support multiple IP packets in one data field.

Yes, it seems mandatory because the size of the Data Fields can be very
large 
(until 58192 bits) compared to the IP packet length.

> 
> A not so obvious one is the support for packet segmentation.
> For me, I would be in favour of making this optional. When 
> you have large data fields, you loose at maximum 11999 bytes 
> (8*MTU-1 bit). 
> But of small data fields this is quite important.

"11999 bits", it's better :-) 
Intuitively, on average, without segmentation, you should loose about half
of the 
average length of an IP packet per Data field. For short Data Fields, it 
can be very important. 

> Of course, the delay could have important consequences. 
> Imagine the scenario where you transmit the first part of the 
> packet in one data field. In the next scheduler slot, the 
> user is experiencing bad channel conditions and if you 
> transmit in the same data field, the user will receive the 
> packet corrupted. 
>
> This would mean that by system design, the 
> segmented packets would need to be outside the mechanism that 
> decides the appropriate Coding and modulation rate.

I think that the segmentation is indeed one of the things that should be
evaluated by taking 
into account all the parameters, including delays, probabilities of data
field corruption, .... 


Jerome Lacan