On Wed, 29 Oct 2008 10:01:31 +0000, "Paul Mansfield" > > I think you mean "asymmetric" rather than "half duplex".
Hi Paul. I do make mistakes, but I did in fact mean to say BOTH asymmetric AND half-duplex. In other words: ADSL is Asymetric AND ALSO half-duplex SDSL @1.5mbs is Symmetirc but only HALF-duplex ON THE OTHER HAND, a T1 is fully symmetric BUT it is also FULL-duplex. In other words, A 1.544 mbs T1 can actually transfer 3.88 mbs - 1.544 in EACH direction. Compare that to SDSL @1.544 mbs which can transfer at most 1.544 mbs total. 1.544 up OR 1.544 down OR some combination thereof where up + down <= 1.544 mbs. People who talk about full-duplex SDSL are usually talking about 768k (or slower) SDSL which CAN be full-duplex (because the 1.5 mbs data link layer can divide its total wire line capacity between the 2 legs) Again, to my original point, someone with an expensive 1.5mbs sdsl circuit can only be assured of proper shaping if they always cap their bandwidth at 768k. Otherwise their traffic shaper would naturally give for example 1.2mbs to an outbound stream, fully expecting easily meed the demand of .9 mbs to an inbound stream. In fact it will be dropping 600k of RTP packets on the floor because it's over capacity. See what I mean? I have experienced this phenomenon first hand, and it is worthy of attention. YES, I could limit the shaper to 768k in each direction, but that means one has to leave HALF of your available bandwidth sitting idle to be confident that the shaper does not make assumptions that result in dropped RTP packets. I see that as a problem worthy of attention, especially as VOIP and other RTP traffic continues to gain traction as killer apps. I'm hoping you developers out there will weigh in on this. Keep up the great work! -Karl --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
