The shallow survey depth of K=18.4 seems a sensible match to POSS II, as typical galaxy colours have B-K~4, and the plates reach B~22.5. From the latitude of UKIRT one can see roughly 30,000 sq. degs, three quarters of the sky. Allowing for likely inefficiencies and weather (see later), such a complete JHK survey would take about 1200 nights, i.e. around 5-10 years depending on what fraction of UKIRT time we use. Such an IR sky atlas would be an incredible resource and in the long term may be a sensible goal. But in the medium term, it is not the best strategy, for several reasons. (i) Much of the exciting science we would like to tackle is much deeper, while still requiring large amounts of data. (ii) Just as Schmidt era surveys fed spectroscopy on 4m class telescopes, we should be looking towards feeding spectroscopy on 8m telescopes. This suggests that we should be looking for exposures of a few hours rather than a few seconds. (iii) Multi-wavelength identifications - from current and imminent X-ray, submm, and radio facilities - are expected to be several magnitudes fainter than the Schmidt limit. (iv) For given science goals, even very ambitious ones, we don't necessarily need all that area, or quite so many sources. (v) Any given area should ideally be covered many times - for reliability, for variability, and for proper motions.
All this argues for looking at the science drivers and tuning survey design accordingly. Of course these are hard to get right in advance, looking 5-10 years ahead, and different goals may push us in different directions. The net result from debate amongst our consortium is to propose not a single monolithic survey, but a survey programme, made up of a suite of distinct component sub-surveys with a variety of areas and depths. The particular areas and depths are driven by specific science goals. However, somewhat inevitably the general result is a kind of logarithmically spaced ''wedding cake'' plan with wide-shallow, medium, and narrow-deep surveys.