One of the interesting features of TNG is the fact that archiving has been envisaged as embedded in the end-to-end operations scheme of the Galileo observatory . This allows the TNG archives to be usable both from a scientific and a technical point of view. Requirements for the TNG archives have been set in ; additional documents on the TNG archives [6,14] are available for reference.
A section of the TNG archives will be hosted at the telescope. This facility may be accessed to gather information on instrument response, trends, calibration procedures, or to retrieve ``standard" calibration files. Access to the archives at the telescope will be restricted to observers and observatory staff, although there is no technical obstacle in allowing a wider access.
Figure 3: Querying the archive: constraints are specified in the fields of an HTML ``form".
Figure 4: Retrieving information from the archive: data are displayed using the HTML tabular format.
The concept of building the TNG archive as a part of the data handling chain had already been formulated in 1992 [15,16]. The archive system at the telescope has now been designed  and is currently under development. The information on instruments will be contained in the Technical Archive, and the calibration files in the Calibration Archive. The ``standard" calibration files stored in the Calibration Archive at the telescope can be used to perform a suboptimal reduction of the acquired files ( Quick--Reduction) to evaluate the scientific quality of an exposure while taking the following one .
All archives at the telescope will be filled directly on the mountain by the Handling and Archiving Tool (HAT) and may be accessed by the DAta Retrieval Tool (DART). The rôles of the HAT and the DART are depicted in Figure 2.
The user interface to the archives at the TNG will be part of the DART, and will be based on HTML browsers. The mechanism is to use HTML forms to prepare alphanumeric strings following the SQL protocol; the strings are passed to the database management system (DBMS), which will handle the query, retrieving database information and displaying it on the window of the HTML browser used. Mechanisms have been envisaged to display graphically the results of a query as X--Y graphs, or to display 1-D or 2-D quick-look (compressed) data. Such concepts have already been successfully tested on prototype implementations [18,19].
Currently, the handling and archiving of TNG data are being simulated using NTT EMMI data . The DART application allowing access to the simulated on-line archive is currently being developed: examples of query and retrieval of archived EMMI data are shown in Figures 3 and 4, respectively. Queries are implemented by filling fields on HTML ``forms" with the constraints the query has to fulfill, while the results are displayed in tabular form. As an exercise, Figures 3 and 4 have been obtained fully exploiting the table specifications of the HTML 3.0 proposal version, currently supported by Netscape 1.1b3 and, with quite a few limitations, by Mosaic versions from 2.5 up.