RE: [lwip-users] TCP problem

[Simon Goldschmidt]

> Is it worth adding a check like other checks that are in lwip_init?

Up to 1.3.0, it seems to have been working with WND == MSS. And although this 
is not a good idea regarding performance, there are applications where it makes 
sense, e.g. if your system is too slow to receive two segments in a row without 
a big gap in between. So adding a check including #error might be annoying in 
this case.

However, I a) don't know if other TCP stacks handle this well (problems with 
remote peers may arise) and b) I guess WND < MSS might not work at all, so at 
least checking this in lwip_init might be a good idea.

After all, I'm still not sure that the change in behaviour seen from 1.3.0 to 
1.3.1 (can be seen quite good in Jan's wireshark dump) is OK (unless it can be 
turned off by setting the window update threshold to zero).

Simon

> 
> Bill
> 
> >Alain Mouette wrote:
> >> May I suggest that a comment about this be added in the config file.
> >>
> >> A special page on the wiki about configuring the many buffers in LWIP
> >> would be awsome too... This is a very obscure area in lwip config :(
> >>
> >
> >Please have a look at opt.h (and it's in there for a while now):
> >
> >/**
> > * TCP_WND: The size of a TCP window.  This must be at least
> > * (2 * TCP_MSS) for things to work well
> > */
> >#ifndef TCP_WND
> >#define TCP_WND                         2048
> >#endif
> 
> 
> 
> 
> 
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users

-- 
Jetzt kostenlos herunterladen: Internet Explorer 8 und Mozilla Firefox 3.5 -
sicherer, schneller und einfacher! http://portal.gmx.net/de/go/chbrowser