Description of Problem:
The timeouts for pppd and pppoe that are assigned by internet-druid interact
negatively and insure frequent disconnects. The timeouts cannot be altered
through either the internet-druid or neat interface.


Version-Release number of selected component (if applicable):
redhat-config-network-1.0.0-1

How Reproducible:
100%

Steps to Reproduce:
1. Configure an xDSL interface with internet-druid.
2. Establish a connection using rp3 or adsl-connect. Be sure to use the
configuration file created in step 1.
3. Don't use the network for about 20 seconds (might take longer).
4. Use "ps afx" to locate the adsl-connect, pppd, and pppoe processes. After
some time, the process IDs for pppd and pppoe will have changed. If using the
RP3 applet or another network interface monitor, the green "connected" light
will turn black.

Actual Results:
After periods of inactivity, the connection will be dropped. It usually takes 20
seconds, but could take longer.

Expected Results:
The connection should be stable.

Additional Information:
internet-druid gives the same value to both the pppd echo interval and the pppoe
timeout interval: 20 seconds. This causes a race condition where pppd issues an
LCP echo request at the same time that pppoe decides to die because of
inactivity. To avoid this problem, the pppoe info page advises against making
these values the same, and suggests that the pppoe timeout interval be 4 times
larger than the echo interval.

The relevant values in /etc/sysconfig/network-scripts/ifcfg-(whatever) are:
  PPPOE_TIMEOUT=20
  LCP_INTERVAL=20

Changing the PPPOE_TIMEOUT value to 80 solved the problem.