Thanks for your response. A few comments, and then a question:

* Looks to the AP just like a standard client, so it interoperates with all APs

What does? A 1200 in WGB mode? If so, then yes it looks like a standard client to the AP but Cisco explicitly states it will only associate with other Cisco AP's. It is not compatible with "all APs".

If you mean a non-root-bridge looks like a standard client to an AP, that's true but the AP must be in "AP with wireless clients" mode which is not one of the test cases I was asking about.

* No configuration required

Of the WGB? Mine need configuration: SSID, encryption keys, etc. It's just like setting it up as an AP except for the mode you select.

Of a (non-)root-bridge? Mine have needed configuration in that mode, too. Same as above.

Not sure what you mean by "no configuration required".

# Disadvantages:

* Supports only a single wired client

Nope, in WGB mode pre-1200 units supported up to eight wired devices. 1200's and later support up to 250 clients. Cisco docs are clear on this.

In bridge mode I'm not aware of any wired client limits. The device is simply passing traffic between the two interfaces.

* WGB itself is not manageable

Not sure what you mean by "not managable".

Back to my question: What is the difference between AP+client mode vs. root-bridge+non-root-bridge mode? Does the AP+client mode make more decisions about traffic routing or prioritization, while bridge-based operation is less "intelligent"? Something else?

My tests show no discernable difference in throughput, even with multiple clients/nonroot bridges. When multiple users consume bandwidth simultaneously, it is shared among them just as you'd expect.

So... what ARE the differences, as far as the decisions IOS is making, between the different modes?

Thanks!

OK, that's good info.

Any idea of the differences in the routing/bridging portion of IOS? For example, when in bridge mode does the 1200 series act like a true (transparent) bridge between the two media, or is there still some semi-intelligent routing decisionmaking happening - and if so, what?

Is bridging any less consumptive of CPU resources? Any more efficient? That sort of thing.

There is no routing functionality at all. If you had multiple VLANs/SSIDs, you'd need an external L3 device to do any inter-VLAN traffic movement (the Fast Ethernet interface does support 802.1q encaps).

I never measured it, but I am inclined to believe that a simple p2p bridge would be somewhat less CPU / resource intensive because the device does not have to poll or track multiple clients ... it only needs to encaps and forward the traffic.

FWIW

I don't mean "routing" in the L3 sense, but the versions of IOS that run on these boxes do support the "bridge irb" command. That's Integrated Routing and Bridging. There's plenty of IOS commands they *don't* support, so the fact that IRB is there means something is going on....