> On Oct 16, 2022, at 2:00 AM, Ian Swett <[email protected]> 
> wrote:
> 
> 
> I have mixed feelings about fixed constants in general, but I agree that 
> referencing the 60 second max in RFC6298 seems reasonable.
> 
> I'm not sure it's worth an errata, but if we were still in the editing 
> process, I suspect we would have accepted a sentence mentioning that.
Maybe, but I would like to read an analysis of this timer strategy from first 
principles. The 60 seconds limit is awfully long in practice, given for example 
lifetimes of NAT mappings. It is counterproductive if packet losses are due to 
physical layer issues, or maybe under DDOS conditions. Exponential growth of 
timers may be justified by congestion considerations, but even that would 
benefit from further analysis.

-- Christian Huitema

Reply via email to