Roaming WLAN / GPRS

Obnoxio The Clown writes "The Register has an article on breaking technology which will (theoretically) allow roaming between WLAN and GPRS (and presumably 3G when it gets here)." At long last, I'll be able to delete my spam from everywhere!

Comments

3G?

[Max Romantschuk]

I don't know about you guys, but I feel like it's more likely that this kind of technology will become 3G than the third generation mobile networks themselves.

Re:3G?

[Michael Hunt]

802.11 as next-generation cell? Unlikely. The contention arbitration mechanisms aren't there, and the protocol was never designed to be used over a distance of greater than about 500 metres.

People like Karlnet (with TurboCell) have tried to push this envelope, largely unsuccessfully. A company I was involved with (who shall remain nameless, as I still own some of their stock) tried to push these technologies as 'long distance wireless multipoint technology' and it plain sucked at it. Square peg, round hole.

Comparing 802.11(a,b,g, or even TurboCell) with 2.5G/WCDMA/GSM/whatever is like comparing shared 10 megabit ethernet with a DS1. The latter has fixed timeslots to avoid two, three, or 30 devices contending for the available bandwidth. Each device is allocated a section of spectrum (bandwidth in the case of a DS1.) 802.3 has CSMA/CD which guarantees that all devices can continue speaking in the presence of collisions. 802.11 uses an even more perverse scheme referred to as CSMA/CA (collision AVOIDANCE,) which avoids collisions in the unlikely event that all nodes can see each other.

Incidentally (I will get to the point eventually, I promise,) cell works by dividing the available spectrum into timeslots, much as a DS1 does. The three most common division mechanisms are FDMA (frequency multiplexing, in which case every device gets its own unique, but very small, section of spectrum), TDMA (timeslot multiplexing, in which all devices contend for the same spectrum, but the timing guarantees each device uses a given slot,) and CDMA (code-based multiplexing, i'm honestly not sure how this works. Qualcomm own a patent on much of it, paid for by the American taxpayer.)

Since each device is guaranteed, deterministically, not to step on the toes of any other device, the devices don't have to be able to see each other in order to avoid major contention issues.

Now, back to 802.11 and CSMA/CA. The CSMA/CA algorithm guarantees that a given device won't hog the spectrum when it can hear other devices contending for it. If you move some of these devices such that they can't hear each other, the theory has it (which is why you have a wireless net running in infrastructure mode with a base station) that the base station will repeat communications from one node to the others. However, node A won't have any way of knowing that node B is after some spectrum, and vice versa. End result, the node which is closer (think physics and speed of light) will get the lion's share of the bandwidth.

Based on these assumptions, as well as the fact that 802.11 operates in spectrum used by microwave ovens, obscure military devices (the UNI/I 5.8 gig band,) and some satellite stuff (sideband interference 100MHz up the scale), 802.11 will never replace a properly thought out deterministic voice-grade protocol.

Now, *MDS on the other hand....

Re:3G?

[Michael Hunt]

Handoff between base-stations would require one of two things in order to work:

a. the 802.11 infrastructure was all bridged, possibly with some interesting VLAN setups (although this would largely be useless, as you don't win anything by VLAN'ing on a shared network)

b. device advertises its address as a 32-bit prefix to each base station it encounters, which then redistributes this into the network's IGP. This would not scale very well either.

An ATM-based wireless protocol would sort out these issues as your IP connection would be an SVC (most likely) which could be handed between switches via PNNI. If convergence comes, it will be from this direction.

Wireless ATM would be a really rad usage of ATM, too. Phone calls and data calls would all be configured as SVCs to very interesting NSAP addresses, and the base stations would have to have full ATM switching logic (possibly with MPLS built in) in order to cope.

Bandwidth

[rf0]

Cool now does that mean instead of paying for airtime on my mobile I can route my calls via my ADSL using VoIP + right software? Now that would be cool and save money

Rus

Welcome to Europe...

[MosesJones]

Anyone who thinks the smart chip is credit card sized is a moron. My mobile has a smart card inside, its the same size as the smart card in my credit card... BUT NEITHER OF THEM ARE THE SAME SIZE AS THE CONTAINER (i.e. credit card or mobile).

Sheesh, you can get smart cards that need all that space for extra memory, but most are tiny things with the external contacts making up most of the support.

The mobile IS the smart card device that can be carried around as it has all of the required elements

1) Contains a smart card
2) Able to interact with other devices over multilpe mechanisms (GSM, IrDA, Bluetooth etc)
3) Smart card can be replaced as require.

Smart cards are NOT credit card sized, that is the plastic that holds them. Its sort of like saying that starter motors are car sized.

Re:Break through technology.... or same thing ?

[caluml]

Or even better - look into Mobile IP. http://www.ipunplugged.com/ is a box we are trialling now. Uses care of addresses, and IPsec. (Oh, and our 3G works nicely ;) )

Re:Break through technology.... or same thing ?

[chotahead]

yah, right. like mobile IP (v4) fixes everything.

how about triangle routing to the correspondent node ... this will make big fun for anti-spoofing filters. you can always reverse tunnel, but what if your home agent is miles away and you're talking to somebody close by? and what about somebody spoofing your binding updates? and you can forget about dynamic handoffs. blag. i don't like MIPv4.

and good luck trying to use ipUnplugged stuff without their Roaming Server, or with home agent and mobile node behind different NAT devices, or trying to maintain advertised functionality without end-to-end ipU boxes.

now, MIPv6 is another story entirely ... ;)