The NASA Space Physics Data System (SPDS) in its present form was organized at a community meeting held at Rice University, June 1993. The SPDS is one of four data systems supported by NASA to support curation of and access to archival data sets relevant to the supported disciplines. For reasons discussed below, the representatives of Solar Physics research at the organizational meeting elected to construct a grass-roots system of pointers to online data sets, focused on the host solar.stanford.edu as being very well known throughout the research community. Although SolarData was originally contemplated as an FTP service only, it quickly became evident to all involved that the World Wide Web provided the perfect expression for this service.
The combination of the traditional simple services involving research infrastructure support with more research-oriented data services has led to the creation of a consolidated Solar Information Center. Here we describe the various components of that Center, both existing and planned.
The Solar Information Center, including all features described here, is available for browsing, trial use, and comment at the URL http://solar.stanford.edu/.
SolarMail was built without formal support simply as a convenience for the builders, and has been maintained with a minimal level of funding. Initially it used a simple homemade database similar in structure to a unix passwd file and updated with an editor. The database was subsequently updated to a hash-table format using ndb, in anticipation of opening up the database to user maintenance. (So far, the maintenance has always been handled by mail requests addressed to a postmaster.) Public database maintenance has been slow in coming, although test interfaces have been available for several years. It was originally designed as a captive login shell, but we have since recognized that the World Wide Web is sufficiently available to the vast majority of users and provides exactly the tools that are needed. In the meantime we have decided to use an ASCII table-based relational database (/rdb), and are in the process of converting once again. We expect to publicly unveil the Web-based interface to SolarMail within the next few months and to gradually wean the system away from maintenance by a postmaster.
Associated with SolarMail are several obvious features that have been provided through mail to special accounts but that lend themselves much better to Web services. These include provision of the directory of SolarMail registrants, real mail addresses of the users, and more general information services, such as postings of meetings, jobs, and other announcements, and distribution of a newsletter. These are discussed below. One other feature that has been provided is the construction and maintenance of mail exploders, an easy place to do so, since everyone on a list has the same virtual host address.
Control of SolarMail has been deliberately loose: participation is basically based on self-identification with the field of Solar Physics. No one requesting an alias is turned down, and the system is not abused. As an extremely well-known host it is the occasional target of pranksters. We exercise a moderate level of security and do our best to keep the host walled off from other hosts on the net, and have seldom had serious trouble. The host is clearly ideal as a World Wide Web server due to its name recognition, its dedicated nature as a community server, and the community feeling of propriety in it, a feeling that should be subtantially enhanced when users are given the ability to maintain their own entries personally.
Given the the necessity of dividing already minimal resources among four distinct groups, it was evident that the only kind of useful ``data system'' we could build would simply be a meta-data system, that is a set of pointers to data already available. In this respect, the solar physics community was fortunate in that a number of observatories had already taken the initiative in placing their data on the net. The representatives of the community at the SPDS organizing workshop decided simply to construct a set of Network File System (NFS) mount points for existing anonymous FTP servers in a single location, and to encourage other data providers to participate. solar.stanford.edu was selected as the site best known to the community at large and always well served by high bandwidth connections to the Internet. We did develop and experiment with such a system for several months until the explosive growth of the World Wide Web came almost simultaneously to the attention of all concerned and immediately recommended itself as the appropriate medium for this kind of service.
We are implementing ``SolarData'', the Solar Physics component of SPDS on the Web, as a self-registry for URL's, using the same database methods as those being used for SolarMail. We decided to use a self-registry, somewhat akin to ``What's New'' services, rather than attempting to implement a crawler because of the tendency for crawler results to become out of date and to contain considerable irrelevant material. This also takes the responsibility for maintenance of the database away from administrators and places it in the hands of the actual data providers, where we believe it belongs. This also substantially reduces the overhead involved in maintaining the system; we have learned from experience with SolarMail that individual administrative effort is the costliest part of operation.
The database is slightly more complicated than a simple list of URL's in order to facilitate searches. We are attempting to provide a rigid classification schema for data in order to make possible relational queries. There are classifications based on time of observation, region of the Sun being observed, the organization of the data (images, spectra, events, etc.), physical observable (intensity, Doppler shift, Zeeman shift, etc.), and wavelength band. The choices are limited to those provided in pull-down menus, but there is always the possibility of adding a category we have not thought of. There is also the possibility, which we are promoting, of registering and providing theoretical models and calculations, certainly a valid and important data resource. The schema and categories will obviously require fine tuning as people begin to register their holdings.
SolarData has just entered beta release; it is being tested and used by the members of the SPDS committee on Solar Physics and by those data providers who have already participated in the remote NFS/FTP exercise. We hope to publicly announce it very soon, certainly by June 1995 at the latest.
We hope that the services planned, under construction, and already existing will truly provide a useful ``Information Center'' for the Solar Physics community. We welcome any suggestions regarding content, style, organization, bugs, and ideas for other services that might find a natural home here.