Re: [lwip-users] IP reassembly

[Jim Gibbons]

We used LWIP reassembly briefly, and it seems quite reasonable, though 
as you note limited to a single packet.

Ultimately, we found that we had very specialized needs, and so rewrote 
it to meet our needs.  As we did that, we were guided by the existing 
code, which has a very nice implementation of the scheme for tracking 
what parts of the packet have and have not been received.

If all you need to do is tune the max re-assembly size, and if you can 
afford the memory, you should be fine.  If you need to reassemble 
multiple packets simultaneously, you'll have to do some redesign work.  
Before you do that, you might want to look back at your networked system 
as a whole.  Reassembly is a pretty heavy burden for a "lightweight" 
system to bear.  Could you possibly manage the MTUs/MSSs so as to avoid 
the problem?

Ed Sutter wrote:

> One more question...
> It appears that LWIP can handle a maximum of one packet
> reassembly at a time, is this true?
> Ed
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users


-