Re: p1 7.2.4: retrying requests

On 06/08/2011 11:05 AM, Willy Tarreau wrote:
> I have not checked how various products handle the situation right now,
> but I'm already afraid of what we could encounter :-/

Yes, especially when people design queries in ways just for maximal 
page-hits under ambiguity of reasons for 1:1 query per hit ratios. With 
that weight, the evolution of HTTP streams (by persistence) does not 
seem effectual (at least in IPv4).

Maybe solutions specifically for UDP in an HTTP stream is not so 
obvious. OAuth MAC appears like the big toe dip to little more than test 
that temperature of the water. If successful then people may diagnose 
such cascade of streams easier, feel the dynamics, and optionally allow 
the length of such UDP (over OAuth) be maximally determined by 
"Content-Length:". Note that I assumed the possibility of IPv4 UDP 
traffic over IPv6 HTTP streams in this path and gateways can assume all 
IPv4 UDP traffic is paravirtual on IPv6 only networks.

On that wonder, maybe "Pragma:" or "Cache-Control:" helps determine 
other factors for HTTP retries, analytic ratios, structured UDP types, 
and globally quantized HTTP headers (into schema).

Welcome IPv6.


--- ---
Web Development, Software Engineering, Virtual Reality, Consultant

Received on Wednesday, 8 June 2011 20:09:51 UTC