(1) Despite our failure to find (or recognize) "standard candles" or "standard rods" relating L and S or d and , they would allow potentially definitive tests of world models and should not be forgotten. Every advance in instrumentation, every increase in astrophysical knowledge of radio sources should be considered an opportunity for determining world-model parameters. For example, recent high-sensitivity VLBI maps have found milliarcsecond jets with bulk relativistic velocities nearly equal to the speed of light, even in extended steep-spectrum sources whose orientations relative to the line of sight should be random. The angular separations between cores and jet knots in a complete sample of such sources might be combined with their nearly "standard velocities" v c to measure directly the angular-size distances corresponding to the source redshifts.
(2) The source counts n(S | ) are generally better determined than most of the data needed for their interpretation. Little progress can be made at high flux densities because essentially the whole sky has been covered with accurate measurements. The sensitivity limits of the current generation of telescopes has nearly been reached. Source counts below S 10µJy at = 1.4 GHz will be difficult to obtain directly, although some limits to the source counts at fainter levels might be obtained from background measurements. Increasing the sky area covered by deep surveys is still needed to reduce the statistical errors, especially at = 5 GHz, where each map only covers about 10-5 steradians. Overlapping surveys at 1.4 and 5 GHz are badly needed to determine the spectral-index distributions of faint sources selected at these two frequencies.
(3) The weighted local luminosity function at = 1.4 GHz has larger statistical errors than the weighted source count, particularly near the peaks of the spiral-galaxy and elliptical-galaxy contributions, so it contributes significantly to the uncertainty in the evolution function E(L, z). Improvements are especially needed near L 1023 W Hz-1 at = 1.4 GHz to interpret the source count near S 1 mJy and decide whether a "new population" of sources exists. This will require an optically complete sample with known redshifts filling a large volume of space; the UGC sample (Nilson 1973) for example. Radio observations of the same sample should be made at = 5 GHz as well as at a lower frequency so that the local luminosity functions of steep- and flat-spectrum sources can be determined independently.
(4) Redshifts of complete samples of flat-spectrum sources stronger than S 0.1 Jy should be measured to confirm suggestions that evolution peaks at z 2. This may not be too difficult because most flat-spectrum sources stronger than S 0.1 Jy can be identified with fairly bright quasars with strong emission lines. If very-high-redshift quasars are common, they will surely be found in such samples. If they are not, only radio-selected samples may be sufficiently immune to selection effects to prove it.
(5) Optical programs like the Leiden-Berkeley Deep Survey should be extended to sources as faint as S 0.1 mJy at = 1.4 GHz to determine the redshift distribution and host galaxy population of the faint-source population.
(6) High-resolution ( 1") maps of sources fainter than S 1 mJy should also be made to determine their angular-size distributions and improve their radio positions for reliable optical identifications.
(7) Isotropy tests at somewhat higher effective resolutions should be made to detect density fluctuations on scales less than 50 Mpc.