[lwip-users] [lwip] questions & suggestions

[Adel Mamin]

Hello.

I spent some time to study the code for lwIP-0.4.1 and I have several
questions:

1. pbuf structure has a field "ref". What is the purpose of this field and
when I should
   use this field?
2. there are the functions pbuf_chain() and pbuf_dechain() in the code. In
what situations 
   do I need to use them? 
3. lwIP-0.4.1 can't assemble fragmented packets and also it can't split
large packets
   into the packets of smaller sizes (the application should take care about
the size of 
   transmitting frames). Do you have any ideas about this kind of
realization: 
   an application doesn't explicitly allocate memory using APIs of lwIP but
gives it the 
   pointer to the data to send. The data has an unlimited length. In this
case the lwIP 
   constructs all the nessesary headers for this data in the different
memory, splits 
   the data and sends the data not actualy copying it. I see that the code
size for this 
   would be larger. What about the performance? It is interesting to know
your opinion.
4. I think that it would be nice if you can compile your code with
predefined symbols 
   like UDP_SUPPORT and TCP_SUPPORT to allow exclude udp or tcp support in
the 
   executable  module to reduce its size. I believe that in some cases it
would be very 
   helpful feature. 

Adel Mamin.
[This message was sent through the lwip discussion list.]