Hi Hui, the obvious way to handle great client GET load at an intermediary is to set up a cache. This makes sure the origin servers don't get requests from this one proxy more often than with a period of approximately max-age. (Using an observation relationship on the CoAP side, this can be further optimized.)
To me it seems the load shedding mechanisms you describe create approximately the same amount of state as a cache would, with the drawback of introducing unpredictable 500 responses. A cache also already includes the logic you sketch in (2), as a pending cache entry will inhibit further requests. So to me it seems a short discussion of the benefit of and implementation techniques for caching intermediaries would be the best way to approach this subject. A useful contribution would be to dig through the massive amount of literature in this space and select a small subset that is useful for LWIG. (There even is a Wikipedia category: http://en.wikipedia.org/wiki/Category:Reverse_proxy -- most of these do way more than we need here, though.) What do you think? Gr???0?8e, Carsten _______________________________________________ Lwip mailing list [email protected] https://www.ietf.org/mailman/listinfo/lwip
