- a New Strategy
The exchange of educational resource metadata between existing repositories,
is the key to the success of the European Treasury Browser project.
Quite a few other
projects have focused on parallel searching in several databases - most of these
using the Z39.50 search and retrieval protocol. Instead of this approach, the
innovative strategy of ETB is to publish metadata to a robust network, from
which participating repositories will be able to extract metadata according
to their own criteria.
The technical platform of the network will build upon the Network News Transfer
Protocol, NNTP, a stable and reliable technology that has been used for posting
and reading news on the Internet for years. NNTP is thus a well-defined and
well-established protocol with many implementations - both free and commercial
- available for ETB developers. This is true on both the client and the server
A major advantage
of NNTP is its robustness in the delivery of messages across a network of servers.
NNTP handles drops in connectivity and can buffer messages for delivery when
the target machine becomes available again. The use of NNTP in ETB is simply
intended as a transfer protocol for XML metadata messages. It is not intended
that anybody should actually read or post the messages, this will be done by
ETB-specific software only.
Apart from circumventing the potential problems of scale and reliability associated
with the use of Z39.50 this has a number of other advantages. The chosen strategy
will allow repositories to define their role in the ETB network in a much more
flexible way, for example:
. only post a subset of (parts of) the records in their database
. ingest only metadata records within a specific topic, to enhance a specialised
repository (instead of providing a parallel search interface to more general
. add additional locally relevant metadata to ingested records
. maintain local policies of the type of resources that should be searchable
in essence, are related to the principal difference between owning a record,
and having access through a search interface.
In addition, the ETB project includes the feature of a central repository, where
all records posted to the network will be searchable. Each participating repository
will have the possibility of defining a local set of rules, governing the number,
type and content of metadata records to be posted to the network (of course
obeying an ETB common denominator set of criteria), and a corresponding set
of filtering rules, determining what records should be ingested from the network.
Clearly, the local
rules for posting records to, and ingesting records from, the ETB network can
vary significantly, and be based on very different types of criteria, such as:
. target audience
. language of resource
. language of metadata
. knowledge of quality criteria used by specific repositories
. topic of resource
This inherent flexibility is one of the important strengths of the ETB network,
and enables participation on a wide range of levels.
Characteristics of the proposed network include: . Each national repository
connects to a given core server, and uses this server to send and receive new
metadata. In principle, a given client could be connected to any of the core
servers, an issue that probably will relate to the chosen solution concerning
security issues. . The core servers will generally connect to two or more of
the other core servers to provide robustness through redundancy. Currently 3
core servers have been set up for trial purposes, one at each of the technical
partners. It is expected that this will be sufficient to obtain the stability
that we need. Because the amount of data exchanged is thought to be quite low
(compared to what the Usenet NNTP servers carry) it is not important to have
many servers in order to provide all clients with a 'nearby' server in the topological
Next step -
setting up a trial case
The software necessary to connect individual repositories to the ETB network
will be developed early 2001. Implementation of this first version of the ETB
software kit at a chosen set of repositories will be the first true testing
of the network strategy.
In the process
of selecting repositories to participate in trial cases of the ETB metadata
exchange network, it is of the utmost importance to realise the existing needs
and possibilities of existing repositories. To convince advanced, existing repositories
of their benefits of participation in the network, it will presumably be necessary
to include other advanced, well-established repositories.
of the full potential of the metadata exchange network, less well-developed
repositories should ideally also be included. Providing new repository initiatives
with a 'kick-start' could also be envisioned. This would be a relevant test
scenario for the ETB native repository developed in WP10. During this process
of trials, a better understanding of the conditions to fulfil in order to be
a member of the ETB network will also become clearer. Participants in an initial
trial case would probably include repositories hosted by ETB partners.
results of trial cases will also differ significantly depending on both the
scope, i.e. regional/national/international, and characteristics of participants,
e.g. commercial or not, similar or dissimilar metadata structures, sizes, technological
level and skills.