On Mon, 10 Oct 2005 13:25:53 GMT, John Navas
<spamfilter0@navasgroup.com> wrote:

>[POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]
>
>In <j9lik1hlgocas98jjqeb3ucr73va8evd7r@4ax.com> on Sun, 09 Oct 2005 10:43:23
>-0700, Jeff Liebermann <jeffl@comix.santa-cruz.ca.us> wrote:
>
>>... Regardless, in my own not so humble opinion, a 2nd
>>>access point is no better than a repeater, and arguably worse.
>>
>>If the 2nd access point is on a different channel (1, 6, and 11),
>>there's no interference with the main wireless router. Why is a 2nd
>>access point worse than a repeater? Store and forward repeaters
>>retransmit everything heard on a given SSID thus doubling the number
>>of packets floating in the air. Since only one radio can transmit at
>>a time, this cuts the maximum thruput in half (or worse). An access
>>point on the same channel as the main router will also compete for air
>>time but is more selective about when it transmits as it only belches
>>traffic to the connected client radios, not regurgitating every
>>packet.

>This all depends on the type of repeater. A dumb repeater is of course the
>worst case. But in a connection-based system like WiFi, a repeater need only
>repeat traffic where there is a connection.

I've only seen three types of repeaters.

One is extremely dumb and will store and forward literally anything
that it can decode. No filtering by any criteria. Fortunately, these
are long gone. Unfortunately, some repeaters (DWL-900AP+) can
simulate this abomination by using ANY or a blank for the SSID.

The 2nd type is one that filters by SSID. It will regurgitate
anything that it can decode for a specific SSID. That also includes
broadcasts, ARP requests, multicast, beacons, management frames, flow
control frames, and such. The SSID filtering is crude and not very
effective at limiting useless traffic. It would be really nice if
there were additional filters available, such as by MAC address or IP
address, but I haven't seen that. Neither have I seen one that only
rebroadcasts traffic that is connected through the repeater.

If the repeater were a true wireless bridge, which only passes traffic
to devices that appear in a bridging table, that would work the way
you describe. However, that's difficult bordering on impossible
because the repeater only has one bridge port. There's no way to tell
if a MAC address is appearing at an input or output to the bridge as
the input and output are one and the same in a repeater. If the
wireless bridge worked on two different channels, one for input and
one for output, then it would work as you describe. However, none of
the commercial repeaters or range extenders work like this. I have
built what I guess could be construed as a wireless repeater out of
two WAP11 wireless bridges back to back on different channels. That's
what it takes to only pass unicast traffic to the connected endpoint.
Nobody sells those (yet).

The 3rd type is a WDS bridge. These are limited to compatible
chipsets but work very well. All unicast traffic is directed and the
bridge only repeats traffic destined to the other WDS device. That's
because the MAC addresses of all connected WDS repeaters are
pre-programmed into each WDS device. This is the least obnoxious
repeater and operates the way you describe.

>Consider a home WiFi system with access point A and repeater B, with two
>clients X and Y, where X is connected to A, and Y is connected through B. The
>access point A does send and receive traffic for both X and Y, but the
>repeater B need only repeat traffic for Y.

The common range extender store and forward repeater does not work
like that. It will repeat any traffic it can decode for a specified
SSID even if that traffic is not destined to a device that goes
through the repeater.

>With two access points, each would have only one client, and thus A would only
>send and receive traffic for X, but with both on different primary channels,
>two of the available three primary channels would be taken up, which is
>considerably less friendly to other users of the band (e.g., neighbors). Plus
>there is the problem of connecting them together with some sort of wired
>connection, the basic problem that wireless is intended to solve.

That will work with two radios and two channels. There's also no
reduction in wireless thruput as the system is full duplex. Nobody
makes such a product and few (except me) bother to implement it due to
the cost of two boxes. There are several mesh network vendors that
also make multichannel, full duplex, poletop radios. For example:
http://www.belairnetworks.com
They have 3 radios in the box.

>In effect, a WiFi access point and a repeater are like two access points that
>share the same channel with a wireless link between them.

No they're not. There's no input or output port distinction on a
wireless repeater. It can't tell whether to forward a packet based on
its MAC address as in a real bridge. Instead, the repeater has to use
some other criteria to decide whether to regurgitate a packet. Except
for WDS repeaters, they just retransmit anything they decode for a
given SSID. Are you perhaps thinking of a WDS repeater?

>True, the extra
>wireless traffic will slow overall network throughput, but for most home
>803.11g networks the speed loss won't be much (if any) of an issue, especially
>in return for the wireless convenience.

Ah yes. Convenience is a good thing. Incidentally, I've tinkered
with the Dlink DWL-G710 range extender. It has horrible thruput
problems when in the 802.11b compatibility mode. It's bad enough that
the access point has to switch between 802.11g and 802.11b in the
802.11b compatibility mode. Getting the DWL-G710 to switch at the
exact same time seems to be a problem. When I tested the thruput with
a DI-624 the packet loss with an 802.11b client (Orinoco Silver) was
terrible. Turning off the compatibility mode prevented connections to
802.11b clients, but at least sorta worked with 802.11g clients.

Are you sure you still want to recommend using repeaters?

>This why I said that in my own not so humble opinion, a 2nd access point is no
>better than a repeater, and arguably worse. I should have made it clear that
>I was only talking about typical home networking, principally the sharing of a
>broadband connection.

Adding a 2nd access point does have it's limitations but nothing as
bad as the excess traffic, chipset compatiblity, and 802.11b
compatibility mode problems that I've itemized. If the 2nd access
point is on the same channel as the main wireless router, then there
is a possibility of mutual interference. However, if they are
isolated by an aluminium foil back insulation RF barrier, methinks
they could survive on the same channel. I installed a 2nd access
point in a long thin house, with one radio at each end of the house.
The two access points can barely hear each other and do not interfere
much. Works fine. Of course, using different channels is the right
way to do this and totally eliminates the interference problem. The
only downside is that one must run a CAT5 cable between the boxes, or
as you suggested, use power line networking to extend the system.

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