On Thu, 21 Feb 2008 15:52:19 -0800 (PST), toe@lavabit.com wrote:

>I've got a wireless ADSL broadband router at one side of a building,

Make and model of the wireless router?

>but its signal is very poor at the other side of the building.
>Therefore I need some sort of "repeater".

Can you get any kind of signal through the building? If yes, then an
improved antenna on your unspecified ADSL wireless router will
probably help. However, if there are too many walls in the way,
that's probably futile.

Alternatives to a repeater are:
1. Power line networking:
<http://www.homeplug.org>
<http://www.netgear.com/Products/PowerlineNetworking.aspx?for=All>
2. Phone line networking:
<http://www.homepna.org>
3. networking over CATV coax cable:
<http://www.multilet.us>

>Let's say that the network I'm trying to get onto is as follows:
>SSID: MonkeyBrain
>Channel: 3
>Network: 10.9.8.0/24
>Gateway: 10.9.8.1
>
>So far, I've heard of two kinds of technology for "boosting" the
>signal.

Rewind for a second. 802.11 wireless is all done on Layer 2, the MAC
layer. There is no involvement with Layer 3 services, that require IP
addresses. The only IP address that really needs to be involved is
the management IP of the device.

Boosting is not a good technical term. I suggest amplification,
repeating, repeating, regenerating, or something similar.

>Technology 1: You have some sort of wireless device that gets an IP
>address on the MonkeyBrain network. This device then broadcasts its
>own network with a different SSID, different channel, different
>network.

That's a back to back bridge/repeater. It's built from two different
wireless access points (not routers), with the Ethernet ports cross
connected. What one access point hears, is retransmitted by the other
wireless access point.

>Let's stay that the new network is:
>SSID: GorillaBrain
>Channel: 8
>Network: 192.168.1.0/24
>Gateway: 192.168.1.254

Reminder. Repeaters don't know about IP addresses. However, there
are devices that do. For example, a WDS repeater does know about IP
addresses.

>When you send a packet to the internet from this new GorillaBrain
>network, it goes to the default gateway on GorillaBrain, which
>forwards it on to the MonkeyBrain gateway, which forwards it to the
>internet.

Yeah, something like that. Again, the way it gets "forwarded" has
nothing to do with IP addresses. It's all done with MAC addresses.

The one exception is a WDS bridge/repeater, which allows a wireless
access point to simultaneously act as a wireless client bridge, access
point, and repeater.

>Technology 2: I'm not sure how this works, but you can actually have a
>device that just boosts the signal.

Those are mostly tower mounted bi-directional amplifiers. It goes
between your unspecified model wireless router and the antenna. I
don't think your application would be a good fit for one of these.

>You've got the same SSID, same
>channel, same network address.

Yep. Same everything because the amplifier doesn't change any of the
data. It just makes the signal stronger.

>You can communicate in Layer 2
>protocols with the all machines on MonkeyBrain because you trully are
>sitting on the MonkeyBrain network.

Yeah, I guess, whatever that means.

>I don't see how you don't have a
>problem with both devices (i.e. the broadband router and the repeater
>device) sending duplicate signals and taking in duplicate signals.

You *DO* have a problem. However, it's not what you're thinking.
802.11 wireless is half-duplex. You can only send and receive, one at
a time. To prevent collisions, where two transmitters xmit
simultaneously, there is an algorithm and timer included to prevent
this. In a give airspace (within range of your system) only one
transmitter can belch data at a time.

When you use a store and forward repeater to retransmit (repeat) your
data, you have two packets going through the air, one at a time, in
order to deliver one packet of data. Since you can only transmit and
receive one at a time, this cuts the maximum thruput in half.
Actually, it's usually much worse than half because the timing and
synchronization are rarely perfect.

If you decide that a store and forward repeater is the answer, you
should try this simple experiment. Drag your laptop, unspecified
wireless router, and repeater into a closed room and set it up to move
data. Use another computer as a server to simulate the internet.
You'll find that with the repeater functioning, performance will suck,
and downloads will be erratic. Remove the repeater, and things will
work smoothly.

What's happening is by creating an artificially small airspace, all
the 3 transmitters in the room are clobbering and colliding with each
other.

>Which technology do you think is better?

In order of personal prejudice:
- CAT5 wire or fiber to the other side of the building is best.
- Power line, phone line, and coax cable networking is tolerable.
- A directional antenna just might work but tends to marginal.
- A WDS bridge works fairly well, but is messy to configure and
supported on a limited number of access points and wireless routers.
- A simple store and forward repeater sucks but can be made to work.
- The absolute worst is a tower mounted power applier.

>Which tends to provide a
>faster connection?

Fiber optic cable and transceivers.

>Are there any other good ways of going about
>boosting a signal so that it's strong on the other side of a building?

See aforementioned list of alternatives.

>(The signal I'm trying to boost is coming from an ordinary broadband
>router that you'd get from an ISP, nothing fancy).

Is "Ordinary" a brand name or are you embarrassed by your selection of
hardware?

>Technology 3: You want to be on the MonkeyBrain network. You have a
>wireless device that broadcasts an SSID such as GorillaBrain. It's on
>a different channel to MonkeyBrain (let's say it's on channel 12), but
>it has the same network address (10.9.8.0/24).

Won't work. Wireless is bridging and works with MAC addresses. If
you duplicate a LAN IP block at both ends of a wireless link, it's
going to act exactly as if you plugged the duplicate IP's into the
other network. Bad idea.

>When GorillaBrain
>receives a frame on channel 12, it forwards it on to MonkeyBrain on
>channel 3. In this fashion, you could have MonkeyBrain using WEP, and
>GorillaBrain using WPA, but essentially they're the same network.

That reminds me. Many WDS bridges will only do WEP, not WPA. This
allegedly a side effect of the WPA key exchange mechanism. I'm not
convinced that this is true and prefer to think of it as a bug. There
are some products and alternative firmware that support WPA encryption
on WDS, so I know it can be done. Chose your hardware wisely or you
may get stuck with insecure WEP.

>(e.g. if you went into the router statistics on MonkeyBrain then you'd
>be able to see all the hosts that are on GorillaBrain, also all the
>GorillaBrain hosts would get their IP address from the DHCP server
>running on MonkeyBrain).
>
>Is there any device the accomplishes what I describe in "Technology
>3"?

If you insist on using two RF channels, only a back to back pair of
wireless bridges will do that. If you can do it on one channel, it's
called a store and forward repeater.

>Can anyone please give advice as to what kind of "repeater" device I
>should get.

Yes. Don't get a repeater. Run the CAT5 or piggyback on the house
wiring.

>My preference would be either Technology 3 or Technology
>2. I think Technology 1 would introduce unwarranted overhead and would
>add the complication of only being able to communicate via IP with
>machines on MonkeyBrain.

You're going to have store and forward overhead no matter what
technology you select.

>Can anyone suggest devices, and perhaps give me an overview of the
>technologies used?

Start here:
<http://www.smallnetbuilder.com/content/view/86/106/>
There's quite a bit on all manner of 802.11 technology. The reviews
are especially good reading. For example, the WRE54G "range expander"
review is quite informative.
<http://www.smallnetbuilder.com/content/view/24690/96/> (8 pages)
<http://www.smallnetbuilder.com/content/view/29840/98/>

There are really only three technologies.
1. MAC layer repeating with store and forward by SSID (only repeats
packets with a specified SSID on a single channel)
2. IP layer store and forward on top of MAC layer client as a WDS
bridge/repeater.
3. Back to back wireless client bridges.

>And don't be afraid to get technical, I've a decent
>knowledge of datacoms (and also of wave transmission and the like).

I'm not afraid of being technical. However, there's no need here.
Methinks repeaters (and some mesh networks) simply suck and should be
avoided. Note that mesh networks are just a collection of repeaters
and suffer from them collision problems. Some vendors have dual radio
mesh boxes designed to avoid the half-duplex collision problems.


--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831-336-2558 jeffl@comix.santa-cruz.ca.us
# http://802.11junk.com jeffl@cruzio.com
# http://www.LearnByDestroying.com AE6KS