[didier <dgautheron@xxxxxxxx>]

Guy Harris wrote:
> On Wed, Nov 19, 2003 at 09:26:47PM +1100, Ronnie Sahlberg wrote:
>
> > You do not use any color filters.
>
>
> Correct.  Not everybody does, so getting rid of the filter would still
> help.
>
> However, it might be possible, with a summary list widget with a
> callback, to have the color set when a row is displayed - dissect the
> packet when the row is displayed and, if there's a color filter,
> construct a protocol tree, run the color filter, and color it then. 
> (Think of it as "lazy coloring".)
But it doesn't solve the main issue: filters are slow.

Can we tell from a filter if we need to fully decode a protocol?
What about a display attribute in the tree structure, with it we might 
'if enclosed' a lot of expensive sprintf calls.

BTW there's a bug in tcp desegmentation when packets are lost, retransmited:
    %
 63.83     11.01    11.01    13448     0.00     0.00  fragment_add_work
  1.80     11.32     0.31                             tcp_segment_equal
  1.74     11.62     0.30    23063     0.00     0.00  dissect_tcp
it's 20MB  AFP capture.

Didier