The key is finding an Ethernet solution which has a dev side and a volume side. I need to get prototyping VERY quickly and easily, but I also need a path to low-cost volume production.  Here is a survey of some of the highly integrated solutions out there.

Lantronix makes modules which, from reading Making Things Talk, seem pretty easy to use, if a bit limited. They are very expensive.

Xecom makes serial-Ethernet modules which are not too well documented, and not yet available through their online dist; but they do have them priced, at $25/1. No path to lower cost, but the price is right.

WizNet makes serial to Ethernet modules which are also poorly documented, but affordable ($30). Here, though, they have a path to production, in their single-chip solutions, about $5. The real problem facing this company is the terrible documentation, and lack of a US user community. Fred Eady wrote a bootcamp article for WizNet’s Circuit Cellar contest, but even this makes it sound like the part is hard to use. WizNet’s best option for quick eval is an AVR-based dev system with LCD, micro, W5100 chip, and interface glue; $200. Or I could buy one of their simple network modules, and use the AVR source code to convert to develop code for another platform, such as the Jennic Zigbee microcontroller.

Rabbit Semi have been making Ethernet solutions for years. From reading the books out there, seem easy to use. Expensive. An Ethernet-enabled module for installation on a customer board is $69/1, $50/1k (RCM4010). It uses their 4000 microcontroller, which has built-in Ethernet, so there may be a lower-cost production path there ($10/1k). The onboard micro would also handle the gateway application. Free dev tools. Dev kit on sale right now. TCP/IP stack in application code, as compared to the WizNet solution which handles most Ethernet and TCP, UDP in hardware. The strength of this company seems to be their emphasis on support, at the obvious expense of cost.

So, how can I best create a gateway system which is flexible, modular, easy to prototype and program? Conceptually I would like to see a central generic controller, and two coprocessors, one for Internet and one for Zigbee. However, everything I have found so far is something of a hybrid situation. I guess the closest would be the EM260 Zigbee coprocessor, and the W5100 Ethernet coprocessor. Though both of these systems rely on a lot of help from the central unit for their operations. I think better, at least for getting started, would be complete coprocessors, that take care of everything for their networks. Then I pass data in, receive data out, translate and buffer data, etc. in the middle.

Rabbit uses a XBee module for their demos (which I believe is in turn is an EM250 with lots of code to make it more plug and play). This type of approach is possible, using a Rabbit board, XBee or similar, and doing the bulk of the processing in the Rabbit. Seems to lack the production path I wish for.

Beyond Logic argues the merits of the Wiznet et al hardware TCP/IP stack. They also look at the Lantronix modules built into Ethernet connectors, and mention a similar series from Digi.

The DigiConnect series has sevral interesting flavors. One is programmed for plug and play, a complete server in an Ethernet connector; another is programmed for development with Net+OS royalty-free RTOS and Eclipse-based IDE; a third allows programing within .Net framework using C#. All three flavors of modules are $50/1.

Another interesting aspect of this system is that the wireless version ($130/1) is footprint compatible with the wired flavors, and also sealed with a fixed antenna for agency approval out of the box. The Net+OS dev platform ($500) includes a wireless version.