[Tomas Kukosa <tomas.kukosa@xxxxxxxxxxx>]

> > |That will, I think, work for TCP - but for other protocols, where the
> > |reassembly is done above the layer that handles the reassembled data
> > |(e.g., IP fragment reassembly), that won't do it; we'd have to handle
> > |that in some other way.
> >
> > Couldn't we add this to the generic reassembly code?
>
>
> No, because the generic reassembly code doesn't know what protocols will
> show up above the layer at which it's doing the reassembly - even the
> dissector *calling* the generic reassembly doesn't necessarily know
> that, if heuristic dissectors are involve.

I have an another idea.
Could not we create a new function
void add_new_data_source_reassembled(packet_info *pinfo, tvbuff_t *tvb, char *name, 
fragment_data *fd_head)
which would save to packet_info together with the new data source the list of frames used 
for its creation.
The save routine could check that list and could save all necessary frames (it should be 
save dialog option).

Any comments?