[j-nsp] ISIS overload bit settings

[Hannes Gredler]

On Mon, Sep 29, 2003 at 11:40:04AM -0600, Danny McPherson wrote:
| 
| On Monday, September 29, 2003, at 10:23 AM, Daniel Roesen wrote:
| >
| >Sorry, I have to clarify: something like "on-startup" exists
| >as "set protocols isis overload [timeout <seconds>]". But there is
| >nothing which watches for BGP keepalives in order to have a heuristic
| >for "BGP converged".
| 
| That's probably a good thing, given the amount of churn in
| the global routing system it's unlikely any full Internet
| BGP speakers receive [post Open Confirm] "keepalives"
| anyways (i.e., if the implementations are optimized such
| that normal updates are counted as keepalives).
| 
| Implementations would be well served by employing the End-of-
| RIB Marker for this purpose (even if Graceful Restart's not
| being used), assuming peers support it, else falling back to
| some reasonable fixed timer (say 120 seconds) or Loc-RIB
| size.

agreed - still you'd need some heuristics to have some tolerance
against faulty states like e.g. "End-of-RIB Marker || Keepalives
received on 80% of all BGP sessions"

/hannes