nd salinity. In this case 
ble to provide a software 
imically linked into the 
| completely specialized 
fic service/provider. 
y, objects resulting from 
able of initiating further 
e a search of an image 
| results object which 
natching the query and a 
would enable a user to 
> service and review the 
of the inventory records. 
JES/APPROACH 
t described here offers 
ne future development of 
providers have complete 
/ish to organize data into 
IS provided type services 
ecide to create their own 
ch and retrieval service 
to replace all data types 
1anagement system (and 
research projects and 
' object databases are in 
1e lifetime of the global 
Rowe, 1986; Haas et al., 
he inclusion of legacy 
ite is only required to 
iishes to support in the 
| minimum set of services 
mple, a site may only be 
insfer, but would still be 
1 open ended approach to 
ieval. Searches can be 
rsite search service, local 
the level of the data type 
| be negotiated among 
e protocols themselves. 
n of results of ongoing 
and schema translation, 
uages might be adaptable 
Jille and Miller, 1993; 
iture search engines may 
telligent searching using 
arth science community, 
ertise "knowledge" about 
xes (Smith et al., 1989). 
no principal distinction 
science metadata (e.g., 
ween metadata and data. 
1 GCDIS and UserDIS 
ns may very well have 
s. The widely differing 
s also will likely lead to 
index, what is data, and 
onable time. Despite this 
hat searches on multiple 
be compared effectively 
by the user, and thus the query process will include 
mechanisms to ensure that the user receives the required form 
of result. The architectural approach shown facilitates the 
introduction of more powerful search strategies in the future 
(Hellerstein and Stonebraker, 1993; Haas, 1989). 
Finally, the concept described above will encourage 
evolutionary and independent development of system 
components. By adopting a fully distributed architecture for 
all components and not mandating the details of the client 
interface and service implementations, the entire user and 
development community can participate in the development 
of components in each of the three layers. For example, 
computer science research may lead to the development of an 
improved intersite search agent. Users can then choose 
whether the new agent provides a ‘better’ service. If it does 
then, over time, it will make other agents obsolete. 
Moreover by establishing a conceptual framework which can 
accommodate the variability of the earth science discipline 
which can guide rather than constrain development of 
components, hopefully minimizing the ‘not invented here’ 
syndrome, it will encourage the development of components 
and support utilities (e.g. APIs) by the entire community. 
Although the architectural concept seeks to strike a proper 
balance between the users' demand for decentralized 
capabilities and autonomy on one side, and complete 
anarchy on the other, a network of the type proposed for 
GCDIS / UserDIS poses significant issues in several system 
quality areas. For example, the accuracy of search results 
suffers as incompatibilities among the vocabularies and 
terms employed by different data providers increases. In an 
unmanaged network, there can be no expectations regarding 
service reliability, availability and response time. For 
example, some sites may respond to a search within seconds 
or minutes, others may not respond for days because the data 
provider experiences hardware problems. 
The solutions to these types of problems are outside the 
scope of an architecture. They depend on the cooperation of 
service providers which, in a network like UserDIS, is 
voluntary. However, the architecture can include measures to 
facilitate the solutions. For example, EOSDIS will not make 
a reliable network, in which all sites are always available, a 
precondition for successful operation. The services will 
provide feedback which lets users judge the quality of a 
response (if they so desire). The. architecture will provide 
mechanisms for characterizing situations where standards or 
conventions exist and are being followed. 
As described above there are several areas where the 
computer science community could contribute solutions to 
the GCDIS and UserDIS challenges. In each area EOSDIS 
will need to pick specific technical approaches which are 
compatible with its implementation time frame, while 
encouraging the computer science community to seek 
improved solutions which can replace the baseline approach 
in the future. 
5. SUMMARY 
The GCDIS / UserDIS concept describes a radical departure 
from the traditional model of data system. By taking this 
concept into consideration in its development of EOSDIS, 
NASA will provide some components of a system in which 
an open interoperability standard can be used to acquire or 
provide data and services, enabling an information system 
to be developed that will operate more as a marketplace with 
37 
positive competition than as a monolithic, monopoly that 
focuses on production and storage of data. 
Such an information system should encourage evolutionary 
and independent development within a single framework on 
an inter-agency and international scale. Indeed its success 
depends on this complementary development. It should also 
provide more flexibility for accommodation of new user 
needs and taking advantage of emerging technological 
developments. Finally, it provides more flexibility to 
respond to the inevitable change in distribution, 
prioritization and funding policies over such a long-term 
undertaking as an earth science information system. 
ACKNOWLEDGMENTS 
This work was undertaken as part of Hughes Applied 
Information Systems contract to NASA for the EOSDIS Core 
System Project (contract NAS5-6000). Important 
recommendations which led to the approach described in the 
paper were received from the National Research Council's 
review of the EOSDIS plans. 
REFERENCES 
Asrar, G. and D. J., Dokken, 1993. EOS Reference 
Handbook, NASA, March 1993. 
Haas, L.M., Freytag, J.C., Lohman, G.M., and P. Pirahesh, 
1989. "Extensible Query Processing in Starburst," 
Proceedings of the ACM SIGMOD'89, June 1989. 
Haas, L.M. et al., 1990. "Starburst Mid Flight: As the Dust 
Clears," IEEE Transactions on Knowledge and Data 
Engineering, March 1990. 
Hellerstein, J.M. and M. Stonebraker, 1993. "Predicate 
Migration: Optimizing Queries with Expensive Predicates", 
Proceedings of the ACM SIGMOD '93, June 1993. 
Lamb, C., Landis G., Orenstein J., and D. Weinreb, 1991. 
"The ObjectStore database system," Communications of the 
ACM, Vol 34, Number 10, October 1991. 
Morrissey, J.M., 1990. "Imprecise Information and 
Uncertainty in Information Systems," ACM Transactions on 
Information Systems, Vol 8, No 2, April 1990. 
NASA, 1994a. GCDIS/UserDIS Study, EOSDIS Core System 
Project, in preparation. 
NASA, 1994b. ECS Science Information Architecture, 
Working Paper FB9401V2, March 1994. 
Ordille, J., and B. Miller, 1993. "Database Challenges in 
Global Information Systems," Proceeedings of the ACM 
SIGMOD'93, June 1993. 
Smith, P.J., Shute, S.J., and D. Galdes, 1989. "Knowledge- 
Based Search Tactics for an Intelligent Intermediary 
System," ACM Transactions on Information Systems, Vol 
7, No 3, July 1989. 
Stonebraker, M. and L. Rowe, 1986; Stonebraker, L. Rowe, 
"The Design of Postgres", Proceedings of the ACM 
SIGMOD'86, June 1986.