【正文】 ows a phone to move between currentlyseparate works while still maintaining connections, as long as both endpoints have IPv6 connectivity. An allIP work emphasizes the distinction between selling access and selling services— a distinction that mobile phone panies currently try to blur. When you make a telephone call, you39。re using their work, and you39。re also making use of the peer agreements they have with other phone works, while enjoying the use of their routing system. Mapping from telephone numbers is much less important than it used to be. I used to make telephone calls by a plicated process. First, I would look up a number in a paper directory or a local cache— either a personal address book or my brain— and then enter this number on a dial (later, on a keypad). By contrast, now I just select the person39。s name from my address book and press call. A recent study claimed that peoples39。 memories are getting worse because they can39。t remember their friends39。 phone numbers. For me, it39。s not just that I don39。t remember my friends39。 numbers anymore— I never knew them. In many cases, a friend sent me a vCard via or Bluetooth, and I never looked at the number field in the first place. Already, the idea of calling a phone (rather than a person) seems faintly archaic. Mobility One of the less wellknown of the family of standards is . 邵陽學(xué)院畢業(yè)設(shè)計(jì)（論文） 36 The point of this standard is to make switching between wireless access points fast. The range of an access point is typically only a few rooms if inside, or about a hundred meters when outside. A person walking around may want to jump between a halfdozen access points in a few minutes. The standard provides a mechanism for doing just that. If you39。ve ever sat roughly an equal distance between two access points, you may have noticed that your connections periodically pause briefly as your working stack decides to switch between the access points. The standard improves this pause time, dropping the switching time down to well under a second. This change makes feasible for mobile telephone use. In an area with a load of .11rsupporting access points, you can walk around while talking on the telephone, and not notice when your call jumps between those points— just as you don39。t notice now when your mobile phone jumps between towers. (Well, in theory you don39。t notice。 in practice, you probably do). This kind of handoff works well if all of the access points are on the same work segment. However, imagine walking down a street full of open access points configured in a mesh (). When you start at one end, you make a phone call and it39。s routed through the first house39。s work connection. As you walk down the street, your call is seamlessly handed off between access points until you get to the far end of your travels. At this point, your call39。s packets are being relayed via a dozen access points to the far end, which isn39。t very efficient. Ideally, before this point you would use the new access point39。s uplink. Unfortunately, each access point has a different routable sub address, so if you switch to the new access point39。s uplink you39。ll suddenly switch IP addresses, which will break your connection. This is where Mobile IPv6 es in. IPv4 also had a mobile variant, which relied on triangular routing— in simple terms, the old work acted as a relay for packets, which added latency every time you moved. With IPv6, IPSec allows the routing tables to be updated securely. Currently, the switching time is around a second, which isn39。t quite fast enough for VoIP traffic, but this time is expected to improve. 邵陽學(xué)院畢業(yè)設(shè)計(jì)（論文） 37 With Mobile IPv6, the boundaries between works bee a lot more fluid. If you39。re sitting in a coffee shop with free WiFi, you may want to use the coffee shop39。s work, rather than paying for use of your cellular provider39。s work. If you walk in while making a call, you would want to switch automatically to using the coffee shop39。s free WiFi rather than the cellular work, which is a more scarce resource and likely to be more expensive. Access Mobile phone panies are very good at charging different amounts for different types of data. Their entire business model is based around this setup. If you39。re paying 10162。 per text message, this arrangement works out to around $750/MB. A lot of people are willing to pay that much for text messages, but at this rate loading the InformIT front page would cost over $500, and I doubt many people would be willing to pay that much for web browsing. You can run an instant messaging client on your phone, and even a protocol as bloated as XMPP works out to be a fraction of the cost per message of SMS. Unfortunately, you can39。t use it to municate with people who don39。t have an IM client on their phone, which reduces the value. You39。re paying not just for the bandwidth for an IM, but also for using the work39。s servers. Telephone calls are also charged at a different rate from that of other data. This design makes more sense, however, since telephone calls have quite different requirements than those of web browsing or . GSM and related protocols use around 12 Kb/s for voice calls. At this speed, you only use around 5MB for an hourlong call— hardly anything on a modern work. Bandwidth isn39。t everything to a voice call, however. If the latency is more than around 200 ms, you39。re likely to notice. More important even than latency is jitter. If the latency increases slightly, there will be a gap in the conversation. If latency decreases, you can only catch up by dropping packets, since you can39。t play back audio at more than realtime speed without noticing. (You can speed it up slightly, but not much.) You can pensate for jitter by 邵陽學(xué)院畢業(yè)設(shè)計(jì)（論文） 38 having a receive buffer at each end, but