tcp/ip assumptions

Mikus Grinbergs mikus at bga.com
Tue Jan 29 00:18:00 EST 2008


Thomas Tuttle wrote:
> I don't understand why you expect the XO to magically figure out how you
> want your network to work.

What I want to know (i.e., to figure out) is how __I__ can make the 
XO work well in my network.  For instance, are there environmental 
variables I can use that will help my setup?  [I do NOT expect the 
__XO__ to "magically" figure out things.]  But it looks to me that 
the way to find out about what is lacking in my network has been 
"try it and see".  I *wish* that topics such as 'proxies' had been 
better described when the XO was released (G1G1) for anyone's use.

> NTP simply doesn't work over an HTTP proxy

I may be mistaken, but I believe I've read Linux descriptions of NTP 
which allowed the server URL to be prepended with a proxy-URL.  I do 
not know whether something like that is supported by fedora (or XO).

> If you want it to work, you have to provide TCP/IP routing to the server!

The reason I called it a "relay" system is because it intermediates 
between my local LAN and the internet.  This system already provides 
several 'servers' for my local LAN, plus several kinds of 'proxies'.

But I also had to define things in the XO, such as RSYNC_PROXY (that 
allows 'olpc-update' over my network by my XO).

An example of a current problem I have is that I have not figured 
out how to provide off-LAN TCP/IP routing to the sugar-control-panel 
specified server from the XO's 'jabber' support.

> There's nothing wrong with the XO making assumptions as to the external
> services available, as the school servers will have NTP servers

But were the G1G1 recipients told "the XO assumes there will be an 
NTP server"?  And were they told how the XO would behave if that was 
not true?  [Note that even school servers may be temporarily down.]


Thanks,  mikus









More information about the Devel mailing list