I'm having a slight design/modeling problem and hopefully someone can
provide me with some good advice or hints.

The problem is that I have a sensor that captures the time period of an
object passing.

The software has a communication layer capable of correctly filtering frames
(just like a packet filter). Each frame has a command code and a fixed
amount of data.
On top of this layer their must be an abstraction of the sensors functions.
This is where the problem arises.

A key problem is to keep the software extendible with minimal future effort.
Currently there exists only a single sensor, which is capable of capturing
from a single source.
A future model would extend this functionality with a secondary channel to
capture a signal from a very accurate time source, including several other
properties.

Probably the most simple way would be to implement each sensor into a
distinct class. This would be simple, but very bad practice. On the long run
it does not work.

My second idea was to provide a fixed interface for a sensor, where each
sensor has different distinct modules. Each module would provides some
functionality, where one is used for the captured signal, a second for the
accurate time source. The hardware should advertise these capabilities and
the software would copy these. This demand only a modification in order to
a) provide new functionality
b) correct problems that arise

- Joris

Joris,

It sounds like you have a number of things that could potentially come
from different sources here. I would have a manager interface which would
expose the different aspects of the input (for example, you have a property
which exposes the time from the appropriate source, and another that exposes
the actual data from the source). This way, you can swap out
implementations easily. You just have to be able to predict where your
boundaries are in terms of what might be swapped out (provider for time,
input, other input, etc, etc).

Hope this helps.


--
- Nicholas Paldino [.NET/C# MVP]
- (E-Mail Removed)

"Joris Dobbelsteen" <(E-Mail Removed)> wrote in
message news:41135d45$0$80310$(E-Mail Removed)...
> I'm having a slight design/modeling problem and hopefully someone can
> provide me with some good advice or hints.
>
> The problem is that I have a sensor that captures the time period of an
> object passing.
>
> The software has a communication layer capable of correctly filtering
frames
> (just like a packet filter). Each frame has a command code and a fixed
> amount of data.
> On top of this layer their must be an abstraction of the sensors
functions.
> This is where the problem arises.
>
> A key problem is to keep the software extendible with minimal future
effort.
> Currently there exists only a single sensor, which is capable of capturing
> from a single source.
> A future model would extend this functionality with a secondary channel to
> capture a signal from a very accurate time source, including several other
> properties.
>
> Probably the most simple way would be to implement each sensor into a
> distinct class. This would be simple, but very bad practice. On the long
run
> it does not work.
>
> My second idea was to provide a fixed interface for a sensor, where each
> sensor has different distinct modules. Each module would provides some
> functionality, where one is used for the captured signal, a second for the
> accurate time source. The hardware should advertise these capabilities and
> the software would copy these. This demand only a modification in order to
> a) provide new functionality
> b) correct problems that arise
>
> - Joris
>
>