Unplugging My Career

Lately I’ve gotten lots of email from ex-clients, friends, and associates inquiring about my recent move to BelAir Networks.

In some way, shape, or form, I’ve been following wireless for a while now. I got my amateur radio license and try to keep up with various spectrum legislation issues, particularly on where I feel our public commons are being threatened. For a while now, I’ve been wanting to shift into the wireless space, and the planets and stars seemed to be in the right order for this decision. In our team, there are approximately 4 “wired” (routing/switching/IP) and 2 “wireless” (RF/cellular), with varying degrees of overlap between all of us. Lento pero seguro, we’re each learning the other side of the fence…

On my first week, I had to revisit a few issues I rarely have to deal with in any great detail.

• RSTP (802.1w): these mesh boxes are, essentially, layer 2 devices with antennas. It’s not uncommon to see hundreds of these things, so spanning tree issues can be quite a problem. Each box has a tiny version of Linux, so from what I can tell, the potential to making switching decisions based on higher layers (7?) is there.
• MobileIP: how you keep a given IP address while roaming around your network.
• IGMP snooping: how our layer two doodads intercept IGMP messages update their MAC tables.
• L2TP: this helps “flatten” different networks for seemless mobility
• 802.11e (QoS): I expected this to be similar to 802.1P, but from what I can tell, it’s not a tag like it is in the wired world.
• 802.11r (roaming): a lot of 11e got rolled into 11r, and I’m still figuring out what goes into making a Wi-Fi <–> GSM handover possible.
• 802.11 in general: I’m quickly learning what’s part of the overall standard, and what’s determined by specific standards or vendor interpretation.

A good example of some of this coming all together is seeing how the sausage is made when it comes to seemless roaming. If you think of it in wired terms, major things go awry when one unplugs a client from one ethernet switch and plugs them in another one: MAC addresses are now associated with different switches, traffic that was en route now has try again, etc. Roaming in wireless is perhaps no different, and when something moves from one AP (access point) to another, traffic en route is disrupted similarly. I searched the 802.11 specification, but cannot seem to find anything that dictates how the client should associate and disassociate from each AP; each client card (firmware, software, etc) seems to dictate everything. One great analogy I’ve heard is that bugs (the “clients”) stick to light bulbs (the “AP”) until they die (i.e. signal fades away), then they begin to look for another AP to associate with.

Up to this point, I’ve pretty much taken for granted what vendors tell me about their IEEE implementation, either in their documentation or in their propaganda disguised as certification. In one week, I’m noted several interesting situations where vendors (not us, of course!) horribly simplify something (”seemless roaming”), is perhaps intentionally misleading (e.g. “full duplex wifi”), or implemented so poorly that it’s intended benefits are outweighed by some other crapiness (e.g. multiple radios in a box interfering with each other).