Talk:TEP-Gateway

From CSL Wiki

1 Thoughts on sharing mote / radio specific information to the PC - Karen

Thoughts on sharing mote / radio specific information to the PC - Karen

There has been some discussion about what the most effective way is to handle the following problems:

A number of different motes are to supported, which in turn use different radios
Some of the different radios require a different radio header
Possibly in the future a single mote may be required to support multiple, different radios

Note that the first two problems can be considered seperately from the third one. Multiple radios on a single mote are not anticipated in the near future, and having to chose the type of radio to communicate at run-time is a more complex problem than having this information specified at compile time.

We want to implement the code to talk to the radio (and fill in the packet header) in a way which is flexible and extensible (since we don't know which other radios may require support in the future), doesn't require too many if / switch / #ifdef statements to call appropriate platform specific functions, and doesn't require significant processing power or memory storage on the mote (since the mote is very limited in terms of both of these).

At some point it seems to be necessary to convert information from a radio-agnostic form to a radio-specific form on transmit, and to do the reverse operation on receive.

The following are the decisions that need to be made:

At which point should the transformation from radio-agnostic data to radio-specific data be done? Should this be performed on the mote, or on the PC?
How should this transformation be done? Is it necessary to use XML, or is a simpler approach acceptable?
How should the mote communicate with the different radios?

[edit]

Question 1: At which point should the transformation from radio-agnostic data to radio-specific data be done?

My initial thought is that it is probably best to do this transformation on the PC. The mote does not have much processing power or memory available, so I think it makes sense to limit the amount of packet parsing work it has to do. The mote could add a tag to all received packets indicating the radio type before passing these packets to the PC. This information would then allow the PC to translate the radio-specific data into the radio-agnostic form.

The major disadvantage of this approach is if there are a number of functions on the mote that require access to radio specific fields in the radio header, extra accessor functions will be needed on the mote. However, since at least initially only one radio will be supported by each mote, this should not require too much additional processing. The only overhead will be writing the extra code to interact with the different radio headers.

[edit]

Question 2: How should the transformation from radio-agnostic data to radio-specific data be done?

As I see it, there are two alternatives to solve this problem:

Use XML descriptions to describe the header contents
Use a simpler set of language specific accessor functions, one for each radio type, that will translate the radio specific header into a generic superstructure, which contains all the information pertinent to any of the different radio headers. This meta-data can then be forwarded on to the application for processing

These two approaches are basically opposite to each other, where the strengths of one approach are the weaknesses of the other. I don't know very much about XML, so the following is like to be erroneous. I hope someone will be able to fix this section soon... My understanding of XML is that it is very well suited for systems that have large numbers of fields and need to support large numbers of different types of objects. Once the system is constructed, it is very easy to extend it

Which approach is best to take is dependent on a number of factors:

How many different radios are currently supported / will be supported in the medium term
How many parameters are provided on a per radio basis
How many parameters are unique per radio (in other words, is our superstructure in the second case going to be carrying a lot of dead fields)

It is my understanding that only two radios are currently supported, and although this number is likely to increase in the future, it will not involve a rapid explosion in numbers. There are also only a small number of parameters supported on a per radio basis, and most of those are common across all radio types (in other words, there are not a large number of parameters which are unique to a single radio). This indicates to me that a simple, host-language style implementation should be sufficient in the short and medium term.

[edit]

Question 3: How should the mote communicate with the different radios?

This question is fairly straightforward in the case where the mote only needs to communicate with one kind of radio. In this case, the type of the radio is know at compile time, so it is simple for the correct set of accessor functions to be compiled in, and they can thus be seamlessly called without any aditional switch statements etc.

However, the more complicated case involves more than one radio attached to a single mote. In this case, the type of the radio can not be known at compile time, and hence must be determined at run-time. As described above, the meta data provided by the PC provides information on the type of the radio to which this packet is destined, as well as information to allow the different fields in the TOS header to be accessed. However, it is important that the transceiver code should not be littered with switch statements (switch on the radio type, call appropriate accessor function) since this is very detrimental to code clarity, and can result in bugs being introduced. Possibly a good approach would be to perform a switch on the radio type at the beginning of the transceiver code, and fill in a struct with the function pointer(s) to be used to utilize this particular radio.

Retrieved from "http://www.lecs.cs.ucla.edu/wiki/index.php/Talk:TEP-Gateway"

Talk:TEP-Gateway

From CSL Wiki

Contents

Thoughts on sharing mote / radio specific information to the PC - Karen

Question 1: At which point should the transformation from radio-agnostic data to radio-specific data be done?

Question 2: How should the transformation from radio-agnostic data to radio-specific data be done?

Question 3: How should the mote communicate with the different radios?

Views

Personal tools

Navigation

Search

Toolbox