Requirements
What the use case needs, in three groups, without saying how to achieve it. The design says how, and Coverage maps each requirement to the design and to the tests.
User requirements
ID
Requirement
Why
From
R-U1
A researcher signs in once, and that covers all services a request passes through
Researchers don't know which services are involved, and shouldn't need to
UC1, UC2
R-U2
Work continues while the researcher is away, for as long as it runs
Workflows outlast any browser session
UC3
R-U3
Everyone with a valid account can sign in, whatever they may later be allowed to use
Otherwise people can't find out how to get access
UC5
R-U4
A service can tell "you aren't signed in properly" from "you have no access here"
A portal can only offer a way in for the second one
UC5
R-U5
A portal can take part whether or not it has a server side of its own
Not every portal has or wants one
UC1
Security requirements
ID
Requirement
Why
From
R-S1
A credential stolen at one service is of no use at any other
Limits a compromise to that service
UC9
R-S2
Services in the middle hold no standing permission to act for researchers
A stored permission is a target, and acts without a request behind it
UC1, UC9
R-S3
Permission that must outlast a request (R-U2) is limited to the one service it's needed for, in time, and can be withdrawn
Long-running work mustn't turn into standing permission
UC3, UC9
R-S4
Each service verifies each request for itself, without relying on what the previous service says
Services are run by different organisations
UC2, UC9
R-S5
Each node decides for itself whom it accepts, whatever services before it decided
Node operators answer for their infrastructure
UC4
R-S6
A service can only pass requests on to services it was set up for
A compromised service can't reach arbitrary others
UC9
R-S7
Requests can't circulate between services forever
A misconfiguration shouldn't take down a federation
UC2, UC7
R-S8
A node can verify which services a request passed through
Provenance can't rest on what those services say
UC6
R-S9
Taking part needs ordinary, standards-based software, without changes specific to one deployment
Some services are outside the federation's control
UC8
R-S10
Use can be attributed to the researcher, and to the portal or workflow system involved
Operators report on it
UC6
R-S11
One request can be followed across all services it passed through
Failures cross organisations
UC7
Privacy requirements
ID
Requirement
Why
From
R-P1
A service learns only what it needs about a researcher
Data minimisation
UC9
R-P2
A node can recognise the same researcher over time, without needing their name or email address
Accounting and audits need a stable reference, not personal details
UC6
R-P3
Credentials and personal details don't end up in logs or reports
Logs are kept longer and read more widely
UC7, UC9
R-P4
Researchers can see what is released about them, to whom
They have to be able to trust the federation with it
UC9