The database can produce, at the end of the night, all the complex FITS files which will constitute the "take home" dataset for the observer. It can also accept at any time during the night quick analytical queries about recent (this night, this month, this year?) history of any meta data parameters, and the results of these queries can be easily presented as various types of plots and graphs.
The observer may wish to keep a "logbook" during the night, in which there is a log entry per science frame, or per every frame; the log entries can be automatically generated and stored from data already collected, and the observer can enter notes from a "notepad" or "logbook" tool easily activated from the control panels. This logbook can then be formatted and output (as paper, mail message, web page, whatever) at any time. Is this logbook strictly private?
The observer may wish to use the database for rapid selection (at any time during the night or day) from the body of data taken during the run so far, according to various criteria. The observer may wish to specify standard text strings, generated from the database, that would be used to caption or annotate screen displays or printed output.
The observer may wish to annotate almost any major database entity at any time. Observers should be able to attach "postit note" comments at will to CCDs, filters, or other documented mechanical/optical features of the instrument and telescope, as well as to their own data, slitmasks, etc.
The observer should be able to request "extracts" of data from the database at any time, in the form of simple tabular files which can be mailed, taken home, etc.
One friendly tool, popped up easily from the standard control panels, should offer interfaces for achieving all of the above "night in progress" database access tasks.
After arriving back at home and starting data reduction, the observer would of course still be able to query the database (through the WWW server) for all public information. Whether observers will have some passworded access to their own private data after leaving the mountain is a design issue. Ideally the answer should be Yes, but reviewers will have to indicate whether normal levels of security for WWW are adequate. Further, standard tools should be provided free for use at the observer's home institution, which enable the observer to perform a useful set of standard analytical queries against the database (perhaps more than are offered to the general public via the web page, because requiring compute power at the local site to succeed?).