Reconstructions of total solar irradiance (TSI) are based on proxies extending back in time which are calibrated against a composite TSI for the last 25 years. Back to 1882 the basic time-dependent information about sunspot areas and locations provided by the Greenwich Observatory, and most recently by the U.S. Air Force operational SOON sites. From these data PSI can be calculated with an uncertainty of <10%, estimated from the comparison with the composite during the last 3 solar cycles. Moreover, the short term MgII index correlated well with PSI and the result can be used to estimate a facular contribution in active regions with an uncertainty of about 20%. However, a major problem is encountered for the estimation of the solar cycle variation due to the network, as represented by the long-term MgII index. Normally this effect is estimated from sunspot numbers. This works very well for cycle 21 and 23, but underestimates the irradiance by more than 30% during the maximum of solar cycle 23. This may indicate that Wolf's definition of the sunspot numbers is finally not such a good estimate for the activity level as correlations with cycles 21 and 22 suggested.