[lwip-users] Re: [lwip] tcp flushing

[Florian Schulze]

On Thu, 28 Nov 2002 11:50:37 +0000 (GMT) Kieran Mansley <address@hidden> wrote:

> On Thu, 28 Nov 2002, Jimi Simpson wrote:
> > If I am correct then this is exactly the same behaviour that I was
> > experiencing yesterday. Basically if you call tcp_output() immediately
> > after the tcp_write() then the data gets sent straight away,
> > irrespective of it's length.
> 
> Yes, looks like our Nagle is a bit too strong!  I'll file a bug against
> this - the sockets interface write() call at least should result in a
> tcp_output(), though I'm not sure if people using the raw interface would
> want it to happen automatically.  Any thoughts?  Personally, I'm
> optimising for low latency so always call tcp_output() from tcp_write().
> 
> Kieran
> 
> [This message was sent through the lwip discussion list.]
> 
> 

I think it should just be documented so people can optimize for either
latency or throughput. It can even change inside the same code. For example
a telnet server should be low latency, while ftp transfers should be high
throughput.
I don't vote for the socket interface, only the raw interface (I think this
is one of the reasons it's called raw).

Florian




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