Originally posted by aztec
I guess I'm confused Puck, before the implementation of the UDP TCP switch I almost never was disconnected. Also if TCP is more reliable why do I experience a drastic increase in text lag, rubber bullets and warps after the switch?
If your connect is "messy"; ie somewhere in your route packets are getting lost, then UDP will switch to TCP. Every one of those lost packets in TCP mode will be resent.
Imagine sending a letter to someone via post office, but you can't put more than five words on a page. You can number your pages, though, so the recipient knows what order to re-assemble the letter at the other end.
The person your sending the letter to gets 1, 3, 4, 6, 9, 12, and 15 (end of letter). They then have to send you a letter saying they need copies of 2, 5, 7, 8, 10, 11, 13, and 14. You then re-send those, but they only get 2, 5, 10, 11, 13, and 14, so they write back asking for yet another copy of 7, and 8. You get the idea. Eventually they'll get the whole thing, but it can take some time.
That's TCP. With UDP you had a letter in 15 parts, but they aren't numbered or ordered or anything else. The recipeint has to figure out how to re-assemble it. If the recipeint gets too confused they can "give up"; that's the point at which TCP takes over.
The problem isn't really TCP. If you have a clean route it will add some overhead and latency, but not as much as you're seeing. The problem is the route to the AH host, which seems to be having problems. In HTCs defense the net traditionally slows down this time of year, and with all the telco bankrupsies outages are neither located or corrected as fast as they once were.
I sympathize, though. I pinged a SpitIX (that's a Spit9, Curval
) from nose to tail with 20mm cannon rounds the other day and he flew off without a care...
