In order for a gravitational lens measurement of H0 to be of primary importance, it is probably going to be necessary for it to have an accuracy ~ 10 percent. Is this realistically attainable? Firstly, we can say that it is possible to measure the time delay to better than seven percent accuracy as demonstrated by the optical observations of 0957+561. This should be repeatable in other selected sources. However, accuracies this good have yet to be demonstrated, at radio wavelengths, partly because variations over the expected time delays have smaller amplitude. It is prudent to take into account such factors as the expected delay (neither too short nor too long), the signal-to-noise ratio in individual flux measurements, and a measured structure function before deciding which few sources to monitor regularly. Simulations should be used for deciding how frequently to measure the flux.
Secondly, we need to be able to model the lenses with a total fractional uncertainty in the predicted time delay also below ~ 7 percent. Here, there are two challenges. As we have discussed, the model itself must be constrained so well that we are confident that there are no other models of the deflector that recover all the observables to within the measurement errors and yet predict seriously different delays. In addition, we must convince ourselves that the perturbative effects of the large-scale structure along the line of sight do not influence our result at this level.
There is a plethora of alternative schemes to measure H0, and each one of them has enthusiastic advocates. In this competitive environment, all methods carry a burden of proof and must demonstrate a reliability and reproducibility if they are to become widely accepted. In the particular case of gravitational lenses, this means that we must derive consistent values of the Hubble constant in several, probably at least four, cases. It is also probably necessary for the models to be specified and analyzed prior to measuring the time delay. The considerations outlined above suggest that this goal is attainable and, as the gravitational lens method is directly physical and free from the calibration uncertainties that bedevil most other methods it is well worth the observational effort to carry out this program.
We acknowledge support under NSF grants AST 92-23370 and AST 95-29170. Support for this work was also provided by NASA through grant number AR-06337.15-94A from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. We thank Ed Turner, David Hogg and the whole CLASS collaboration, especially Ian Browne, Chris Fassnacht, Sunita Nair and Tony Readhead, for discussions.