Re: [lwip-users] Reassembling messages

[Keith Lockstone]

On Mon, 16 Jul 2007 11:03:18 +0100, Kieran Mansley wrote:
> 
> On Mon, 2007-07-16 at 10:56 +0100, Keith Lockstone wrote:
> > 
> > We've got LWIP up and running on a H8S/2378 using the SMCS LAN9215i.
> > However, a system problem has come up that we haven't found a simple
> > answer to.
> > 
> > The datagrams we send have different formats and lengths and to work
out
> > where the starts and ends of the datagrams are is not a trivial
task.
> > What are the typical approaches to formatting a TCP/IP message so
that
> > reassembly is a trivial task?
> 
> We'll need some more detail I'm afraid.  At what level are you talking
> about reassembly?  e.g. is this in your application, in the TCP/IP
> stack, or in your driver?
> 
> I'm guessing you're talking about the application.  Apologies if I'm
> wrong.  Assuming that's the case the normal way to do this is 
> to modify
> your format so that the first word is the length of the message.  Then
> you can read that and know how much to expect. 

It's at the application level - we seem to be faced with a stream of
bytes coming out 
of LWIP without any indication as to the boundaries of the original
messages that 
were sent.  Do we have to add something like SLIP on top of TCP/IP to
define the 
message boundaries, or have we missed something?

Keith.