dlm object growth and storage of past values

[albertotb]

Hi,

Is there any technical reason to store the whole time series used to create a model, besides plotting? If a model is grown by appending new data it can be a huge limitation to store all the values in memory, since it will eventually be too slow to save and load from disk. Besides, as far as I understand it, the model only needs a handful of parameters in order to perform the forward filtering and compute new predictions. Thus, it would be great if we could have the option to create a "minimal" dlm object, that only stores the information strictly necessary to compute basic operations such as appending, fitting and predicting.

Thank you in advance