Archive for November, 2008

Nov 4 results: FCC opens path

Wednesday, November 5th, 2008

On November 4 the FCC cleared the way for the use of TV channel guard bands for I’net access. As TV moves from analog to digital, the need for an empty space, a guard band, between channels to prevent them from interfering with each other is reduced. Getting more TV in less spectrum is the promise of digital TV and this FCC action is an attempt to utilize that promise. It is a second step after auctioning off the upper UHF channels (51 to 80) freed up in the planned transition.

The idea on this use of the guard bands for I’net access is that, while the equipment must still be FCC approved, transmissions do not need a license. There will have to be provision to make sure the equipment does not interfere with nearby digital TV transmissions. This will likely mostly depend upon a location database more than it will signal sensing.

A key feature of using this part of the spectrum is that it is at a lower frequency than the traditional 2.4 GHz WiFi which means it will suffer less attenuation getting into and out of buildings and other solid objects. It will tend to have a longer range with some trade-off for lower data rates.

Joshua Breitbart has a good description in Open the airwaves and the sky’s the limit and why the ‘free is good’ folks are very happy.

Wireless access is not a full replacement for wired connections, but it is a much cheaper way to bring people the Internet. Mobile phones are far more widespread than in-home computers with broadband connections, especially among the groups currently marginalized from the Internet. …

Once you don’t have to rely on big, corporate license-holders to get a connection, you can start to invent entirely new devices and applications. The FCC used the same kind of open platform for innovation with the 2.4 gigahertz band. That led to an astounding array of inventions — cordless phones, remote controls, microwave ovens, and wi-fi routers — all sharing one tiny piece of the airwaves.

Google has been one the primary backers of this idea. Catherine Holahan described their effort in Business Week’s Google’s Plans for the Space Between Your TV Channels

Companies such as Clearwire are providing I’net service similar to that envisioned by Google. This FCC action will make that type of service easier to implement over more area.

eBox: caveats and limitations

Sunday, November 2nd, 2008

As an approach to network administration, eBox is intended for those who don’t want to gain the expertise or spend the time needed to understand all of the nuance and interrelationships between services of a modern office network server. There are two basic approaches to simplifying things like this. One is to handle interdependencies automatically and the other is to depend upon a standard model or template for the system. eBox has a lot of work to do in both of these areas.

The sales pitch for eBox does not clearly indicate that its standard model is a server with two interfaces, each on its own network. In the eBox network settings, the interface with the ‘external’ box checked serves what is usually known as the WAN (wide area network) side that is attached to the I’net. The interface without this ‘external’ box checked is for the LAN (local area network) side. In the standard eBox network model, the eBox machine provides many services that are usually provided by standalone NAT router devices. See the eBox Network Scenario.

In the first scenario eBox is placed between your router and your local network. In the second one, eBox is installed just like another machine in your network. The latter has some limitations as we explain later.

Many micro businesses and home based businesses have a modem that connects to the broadband service, usually DSL or cable, provided by an I’net service provider. This connects to a NAT router that will have several ethernet ports and perhaps a wireless network that services the local machines. These NAT routers can run from well under $40 to several hundred and provide services such as network address translation, firewall, DHCP, and DNS configured by a web based interface. They are standard ‘off the shelf’ appliances that work very well for the basic WAN to LAN interface. If you set up an eBox machine that is a peer under this router with the other local machines, its capabilities become limited due to its model and that model’s necessary interdependencies.

The key here is whether to model is based on the assumption in between the source and destination of network traffic or just sitting in the network providing services to its peers. The eBox model is mostly based on the presumption of throughput.

Services that are strictly for a LAN, such as file sharing and user definition can avoid any interaction with upstream firewalls and traffic controls. Services like traffic shaping and load sharing seem to need to cooperate closely with firewalls yet eBox does not appear to define this interaction for the user. Then there is the proxy service which lists a dependency upon the firewall yet has no obvious need for that dependency. This is a state of confusion noted in the Firewall rework discussion.

The firewall module is one of the most criticised and controversial parts in eBox. Abstracting the firewall configuration is a tricky work, we tried our best, and it seems our approach can be improved.

It appears that eBox is rather young. It does not yet have the service coverage that its competitors like Webmin do, partly due to its primary feature of automatically managing interdependencies between services. This automation feature is perhaps also a reason why flexibility in configuring services like file sharing and web services is limited. A conservative approach to management of interdepencies is also why users may find the feature set restricted in ways that make like difficult for them, especially if not adhering to the network model presumed by eBox developers.