- Information about supports (support name, status, number of copies)
is locally hold by each server keeping it private.
- Collection’s meta-data files are duplicated on all servers,
but revisions are managed by a uniq GIT “master” server.
- Archive’s data files are stored in a cache for each collection by each server.
USER (see Who) query is distributed to all servers.
If none of them can fulfil the query, the server receiving
the initial query asks for the USER’s mail address and registers it.
The USER mail address is never sent to the other servers.
- As the catalog provide static HTML content, a DNS’s round robbin is usable to switch between server’s WEB interface in order to recover from a site crash.