Archived from groups: alt.internet.wireless (
More info?)
Jeff Liebermann <jeffl@comix.santa-cruz.ca.us> wrote:
>On Fri, 08 Oct 2004 10:53:26 -0700, John . <john@notme.com> wrote:
>
>>Jeff Liebermann <jeffl@comix.santa-cruz.ca.us> wrote:
>>
>>>On Thu, 07 Oct 2004 13:42:22 -0700, John . <john@notme.com> wrote:
>>>
>>>>From the DLink documentation I can't determine if the DWL-G810 bridge
>>>>can be used to interconnect 2 wired Lans into one?
>>>
>>>No. One device, one MAC address, and one badly written data sheet.
>
>>Maybe I misstated my question, but wouldn't ONLY ONE mac address be
>>passed between two DWL-G810's, the mac of the bridge? Even though
>>each wired LAN in two separate buildings has 5 or 6 PC's, wouldn't two
>>DWL-G810's just be passing traffic to/through the other corresponding
>>bridge?
>>
>>DWLG810--pc1--pc2--pc3==>switch (all cat5e)
>>+
>>+
>>antennas
>>+
>>+
>>DWLG810--pc4--pc5--pc6==>switch (all cat5e)
>>
>>Does this work?
>
>Nope. The problem is the type of bridge. Again, I'm guessing that
>the DWL-G810+ is a "imple bridge and not a transparent bridge. It's
>difficult to tell from the data sheet. The difference is that a
>simple bridge only deals with one MAC address, while the transparent
>bridge or "network bridge" is designed to connect two networks
>together as in your ascii drawing.
>
>Reminder: bridges deal with MAC addresses, not IP addresses. Let's
>start with a wired LAN bridge. (Incidentally, a switch is a bridge
>with more than 2 ports.) The wired bridge builds a table of MAC
>addresses that it hears on each port. It looks something like:
> Address Port
> MAC1 1
> MAC2 2
> MAC3 1
> MAC4 2
> (...)
> MAC30 2
> MAC31 1
>For convenience I decided the odd MAC addresses would be plugged into
>port 1, while the even MAC addresses in port 2. The 802.3 ethernet
>MAC address header includes the source and destination MAC address of
>each packet. The source address is used to populate the table, while
>the destination address is used to determine which port it needs to
>traverse.
>
>bridging decision time:
>1. If the destination MAC address is in the bridge table, the packet
>crosses the bridge.
>2. If the destination MAC address is in the bridge table, but on the
>same port as the originating MAC address, the packet does NOT cross
>thr bridge (because it has nowhere to go).
>3. If the destination MAC address is NOT in the bridge table, it is
>assumed to be local traffic and does NOT cross the bridge.
>4. If there is no destination MAC address, as in a broadcast packet,
>then the traffic crosses the bridge.
>
>Remember, the whole idea of a bridge is to reduce network (or
>wireless) traffic.
>
>With a two port bridge, such decisions are fairly trivial in that all
>that needs to be decided is whether to cross or not to cross. With a
>switch (more than 2 ports), a decision as to which port needs to be
>made. A bridge or switch also needs a little time to make the decode
>the header and make a decision, so some buffering (FIFO) is usually
>added to each port.
>
>I'm going to ignore the added complexities of a VLAN, spanning tree
>algorithm, and WDS repeaters.
>
>Now we introduce a wireless bridge. Let's start with the "simple
>bridge" and it's one MAC address. The last thing you want is for
>every last lousy packet on your LAN to go out the wireless port on
>your router and pollute the airwaves with useless traffic. The simple
>wireless bridge works exactly the same way as the wired LAN bridge
>exept that it's only concerned about traffic to one computah. The
>access point has a table of radio MAC addresses and makes a decision
>if the packet has a destination at the other end of the wireless
>bridge. Multiple simple bridges are not a problem as the table can
>have multiple "ports" or radios. This is the way all of the common
>wireless access points and client radios operate. A wireless access
>point should really be called a wireless switch.
>
>Just one problem. Each client radio only passes one MAC address to
>the access point. If you build a network of PC's behind the client
>radio, the wireless access point cannot determine if a given packet is
>destined for any of these computahs because the access point MAC
>address table only has one table entry for the client radio.
>
>To connect two networks together, a wireless transparent bridge is
>needed. It's exactly the same as a wired bridge, with its MAC address
>table, except that the table is now duplicated at each end of the
>bridge. Every time a new MAC address appears on one end of the
>bridge, the information has to be added to the table on the other end.
>The protocol for doing this is proprietary by manufacturer, which is
>why mixing different radio types is a bad idea. The size of the MAC
>address tables is also limited in many bridges. The cheapo bridges
>are 31 MAC addresses, while the fancy one's (Proxim, Alvarion) can
>handle about 2000 or more.
>
>So, why can't you just take two access points, aim them at each other,
>and use them to bridge two networks? Well, they can but they lack
>support for duplicating the bridge table entries at each end. All the
>hardware and most of the software is there, but without a bridging
>table protcol, the common access point can only deal with one MAC
>address per radio.
>
>That leaves us with your original question, which reduces down to why
>can't I pile a bunch of MAC addresses behind a simple bridge (client
>radio) and use it to connect two LAN's together though a single MAC
>address? You could if the access point was able to store MAC
>addresses for the destination PC's. 802.11 will encapsulate the 802.3
>ethernet packets. The bridge will see only one of the encapsulated
>MAC addresses, and only one computah will be able to communicate.
>Were it not for the encapsulation of MAC addresses, it would be
>possible.
>
>I'm not sure how the "game adapters" and other bridges handle the
>multiple computahs. My guess(tm) is that they expose the encapsulated
>source MAC address in the 802.11 header, which makes the traffic look
>like multiple client radios. The common access point can easily
>handle multiple radios. Again, I'm not sure about this and need to so
>some wireless sniffing to be sure (yet another project).
Jeff, thanks for your complete exhaustive explanation. It really
really too bad that manufacturers don't more fully explain their
equipment.
My goal was to replace an older pair of DLink DWL-900+ (b mode)
bridges operating in one to one bridge mode. It was fine
interconnected two small wired lans (about 4 PC's each) into one
network neighborhood.
I just replaced the DLink's with a pair of Netgear WGE101 (g mode)
bridges. To get the Netgear bridges to work, I had to set them to
ad-hoc mode with WEP security.
It's working, but I have an uncanny feeling it's not a true
transparent bridge. I could easily see all the shared folders from
both lans, but had trouble with Internet access through the
router/switch that is on one LAN only.
There's very little setup with the Netgear, so it's hard to see what's
happening.
BTW, qcheck throughput went from about 4.5 mbps actual throughput, to
11-12 mbps actual throughput.
Except for the ad hoc mode strangeness, it seems to work.
john