RMSLE as error metric to compare between datasets with different scales?

Question

Is it possible to use the Root Mean Square Logarithmic Error to compare the regression results that are achieved for different datasets?

For example if one dataset ha