Originally Posted by
Raist
Oh boy... this network protocol you speak of is not SE's protocol... it was actually developed by the government---more specifically the Department of Defense. TCP/IP is a standardized protocol, and it is the bulk of what is used on the internet. And yes, in a way... the internet is broken. More accurately, it is buckling under the strain of high traffic along our specific routes.
There is a lot going on here that some of you may not fully understand, as evidenced by that last quoted section. TCP/IP is a common protocol used all over the place---SE is NOT the first to use it for a game, and many have been successfully using it for theirs. Blizzard included. In case you didn't know, normal websites are hosted on TCP (default is port 80), and the data is tpyically transferred through the TCP/IP protocol. Open up resmon or run netstat while on this site if you don't believe me.
The problems come in primarily because of why TCP/IP is chosen. It has it's own system of error detection/correction to ensure information is delivered as it was originally sent. By contrast, UDP (used in XI) has no error detection or correction scheme---packets can be lost completely, or arrive completely out of order. Unfortunately, part of the correction scheme in TCP/IP also adjusts flow control by varying methods, depending on suspected problems affecting the quality of the connection. In short, if it detects enough problems with packet delivery, it can reduce the transmit windows and such and effectively throttle the speed downward in an effort to control errors induced by high congestion. If, after a while the condition doesn't improve sufficiently--it may adjust and throttle things even slower.
So, what we've likely wound up with is a communication problem that the protocols/policies are trying to auto-correct, and not being able to recover properly for various reasons we may not be able to discern, but we CAN see the symptoms of it with some simple tools like pathping, tracert, etc.. Fortunately, there are teams of experts out there employed by the ISP's to do just that--detect, investigate, and correct these problems. The problem is, many have very lax policies for what triggers an alert, which causes situations like this to easily slip under the radar. It isn't happening on a grand enough scale for them to automatically put someone on it--it's isolated along routes to a specific endpoint, but they are not all from the same origin. So, there needs to be extensive isolation in order to properly investigate and correct. That is where we can influence the process. By us reporting the routing details to people more directly connected to these experts in the field when we are experiencing trouble, we are in effect becoming their means of detection. So, they sort of NEED us to submit this data for them to be able to actually see a problem exists that needs to be addressed.