Re: AW: AW: [Madwifi-devel] [PATCH] RFC: link distance specification From: Darrell Budic - 2004-05-26 19:56:57 Attachments: Message as HTML ```On May 26, 2004, at 10:42 AM, Michael Renzmann wrote: > Hi. > > Togg wrote: >> No no, actually I meant the value suggested by Greg in the formula >> (ack/cts=3D(2*slottime) + 24). My experience with long links shows = that=20 >> the old default value of 25=B5s works >> without packet dups (which mean bad throughput) till distance of=20 >> around 400 >> meters. > > Is there anything that speaks against timeouts that are a little=20 > longer than actually needed? Longer timeout values will decrease the maximum potential throughput of=20= a link (more of our available tx time will be used by timeouts), and=20 may slightly raise the latency encountered over our link (packets may=20 have to wait longer to get transmitted). This may get more pronounced=20 the more nodes are using the same AP, since the timeouts are used by=20 802.11x to determine when it's safe to send another packet and have a=20 reasonable expectation of not stomping on someone else when they are=20 trying to tx. On May 26, 2004, at 8:37 AM, Togg wrote: > P.S: can someone explain cts timeout meaning? I only have Point to=20 > Point > links, therefore I think I really don=82t need cts at all? Pls. help = me=20 > with > this question. You probably still do, because cts/rts is frequently used for=20 fragmented packets and some larger packets. It can be used for any=20 packet if you set the driver up that way. Remember that both nodes will=20= be trying to transmit over the same medium, and may need to avoid=20 stomping on each other. You may use it less than a multi-node network,=20= but you probably still use it occasionally. For the full details, get=20 the O'Reilly "802.11 Wireless Networks" book, it has an excellent=20 description of the whole thing. -Darrell ```