Back-end development: responsibilities and practical exercise
Back-end development defines server processing, data persistence and access conditions. It turns business rules into operations with explicit inputs, effects and errors. An interface displaying success does not establish that these operations completed.
These sheets describe activities and teaching exercises. They do not constitute job offers, certifying courses or recruitment promises.
Develop server logic, data exchanges and application rules so actions are authorised, validated and recorded consistently.
Typical activities
Define data models and API contracts.
Check input, permissions, duplicates and processing failures.
Prepare migrations, useful logs and recovery procedures.
Responsibilities to agree with the team
A working API does not prove the visitor journey works. Agree states with front-end developers and deployment and recovery constraints with operations.
A first piece of work to demonstrate
Demonstrate invalid input, unauthorised actions, duplicate submission and recovery after failure.
This is an editorial description of common activities. Job titles and scope vary across teams; it is not a certification framework.
Build a request service that validates entries and maintains a processing status.
Work on a copy or demo environment. Define the data used and the expected result before you begin. Keep your decisions and the difficulties encountered in order to be able to explain the process.
Describe the contract of entries, results and errors.
Check the rights to each operation and define the necessary data.
Test duplicates, limit values and dependency temporarily unavailable.
Prepare migrations, backup and recovery procedure.
Deliverable and checks
Document accepted fields, responses and processing states for the enquiry service. Provide accepted, rejected and repeated request cases, the records created and a way to trace failure without exposing private enquiry content.
Submit invalid data without the form, then try accessing an enquiry owned by another account. Check server-side denial and absence of unintended effects. Also test repeated requests and an unavailable dependency against the documented rules.
A validation in the browser does not replace server control. Messages and logs should assist in the diagnosis without exposing private information.
Collaboration to explore: front-end development.
Fictional deliverable excerpt : The handler checks incoming data, applies the agreed rules and returns a result the interface can explain.
What passes between the roles : Both sides need a shared understanding of fields, accepted formats and possible responses. Checking the journey means examining failure, success and actual enquiry handling beyond the appearance of the button.
Fictional teaching material: this excerpt illustrates part of a deliverable. It is not a client reference, a complete specification or evidence of professional qualification.
Sample excerpt
Fictional operation: create a booking enquiry.
Business request identifier: demo-request-17.
First processing: create one enquiry.
Repeat submission: recover the agreed outcome.
Same identifier, different content: reject or resolve explicitly.
Concurrency test: two identical requests arrive before the first finishes.
Expected: one business enquiry, a retrievable outcome and no second creation.
What the author should be able to explain
Repeated-request behaviour is a design rule; POST alone does not prevent repeated effects.
Define the identifier scope and retention period.
Validate checks and concurrent processing on the server.
Does disabling the button prevent duplicates?
It can limit some clicks, but requests can be repeated in other ways. Define server behaviour, storage constraints and the returned outcome. This is an example contract, not a complete implementation. Also replay the same identifier with different content and retain the rejection outcome defined by the fictional contract. Check the number of created objects, not only HTTP responses. A result lost after writing should be retrievable without triggering another creation.
Technical reference : MDN · POST · . The excerpt is an original teaching example.
Assess the quality of the work
Take a request to change a fictional record and examine what the server accepts for each test account. Visible fields do not define actual permissions. Data validation and record authorization are separate controls that should remain effective when a request is sent directly to the API.
Scroll the table horizontally to read every column. With a keyboard, focus the table area and use the arrow keys.
Assess the quality of the work
Skill
Evidence to examine
Failure to spot
Inputs and business rules
Define permitted fields, formats, bounds and state transitions. Prepare responses for unexpected fields, absent values and stale data.
The server treats a client-supplied value as already validated or accepts a transition impossible in the intended journey.
Object permissions
Describe permitted operations per account and check permission on the targeted record for each relevant request. Use separate fictional objects.
Changing an identifier exposes or modifies another account’s record despite the interface having hidden its link.
Error handling
HTTP status, public message and logging have clear roles.
The response exposes a technical trace or reports success after failure.
Repeated operation
Define server validation and handling of the same request sent twice.
A double click or retry cannot create two business operations; protection does not depend only on the button.
A focused next step
Prepare two test accounts with separate records and replay the same change across account-object combinations. Check rejection, stored effects and public messages without exposing secrets. An error status is insufficient if the operation still changed data; verify the absence of unauthorized effects too.
When one URL produces different responses according to Accept-Language, Vary identifies that cache selection criterion. Keep the value consistent on 200 and 304 responses. Separate language URLs do not require this header merely because the website is multilingual.
If-Match compares a strong validator before a write. Under the illustrated contract, a stale version receives 412; keep the draft to compare changes.
Dates indicate when references were consulted. These facts are neither a site audit nor an overall compliance statement.
A review case to reproduce
A retry creates two operations
Situation to examine
An accepted request is replayed after a network delay. The service creates another operation although only one was requested.
Proposed check
Replay the same test event in a copy and inspect identifiers and writes. Define repeat handling and partial-failure behaviour. Verify the fix without real messages or payments.
Evidence to retain
A test log links event, identifier and outcome. It demonstrates the chosen rule and scenario limits without exposing sensitive data.
Fictional finding example
Two retries carry operation reference DEMO-42. The demonstration log shows two receipts and two writes although only one operation was intended.
Interpret this outcome
Define and test duplicate handling. Two successful technical responses do not justify two business operations; POST alone does not guarantee idempotence.
Three complementary checks
Fictional cases: adapt these expectations to your test contract and retain the limits of the check.
1Identified operation
2Checked write
3Retrievable result
Scroll the table horizontally if needed; with a keyboard, focus it and use the arrow keys.
Scenario variations and expected evidence
Variation to test
Expected result in this case
Evidence to hand over
Two identical concurrent submissions
One operation under this fictional contract
Business identifier and write count
Same key, different content
Explicit rejection or resolution without a second creation
Key, content fingerprint and expected response
Response lost after writing
Retrieve the result without recreating the operation
Retrieved object linked to the first submission
Decision to hand over : Check business effects and storage; the contract must define scope, retention and concurrency.
Extended exercise: Prepare a demo endpoint that receives a request and returns an explicit state. Use fictitious data and a local environment without real sending. Test a missing field, a value that is too long, unexpected content and a double send. Validate the server-side data, even if the form already provides controls. Distinguish the technical acceptance of the request from its treatment by a person. Logs should assist in the diagnosis without exposing the information entered unnecessarily.
What to deliver with your work
Data contract and example queries
Server validation and error responses
Limit and dual shipment tests
Secretless Setup Instructions
Acceptance criteria
These criteria are a working basis to adapt to the context. For each, note the action taken, the expected result, the observed result and the information remaining to be confirmed.
An invalid query does not produce a partial effect.
Size limits and formats are controlled on the server side.
The double sending has a documented and tested behavior.
Demo logs contain no actual secrets or personal data.
Deliver the contract, possible answers and processing conditions with the code. Describe environment dependencies without copying their credentials. A person resuming the service must be able to reproduce an error with a test dataset. Separate local checks from checks that would depend on an external service; Do not report a successful actual shipment when it has not been made.
An operation must remain understandable when it fails or repeats. Compare immediate processing with a deferred task according to duration, dependencies and expected outcomes.
Processing within the request
When to choose it
Suitable for a short operation whose success can be confirmed immediately. Define validation, permissions and errors before writing the result.
What to plan for
Check repeated requests and concurrent changes. An error response must not leave data in an unexplained state.
Deferred processing with a tracked status
When to choose it
Useful for long imports or external services. Return a reference that allows the task status to be retrieved.
What to plan for
Define retries, permanent failures and duplicate prevention. Distinguish received tasks, tasks in progress and completed tasks.
A risk to anticipate
Partial success can be mistaken for a successful operation. Document the authoritative result and the data needed to understand interruptions.
How to decide
Replay a request and make a dependency unavailable. The final result must remain consistent with the reported status and diagnostic records.
Compare processing rules, permissions, persistence and service contracts. A demonstration response differs from a protected operation with failure and recovery cases.
Recurring operation
Specify updates, monitoring and backup checks. Separate operation from functional evolution; an available server does not prove business operations complete.
Exit and handover
Plan code, schemas, migrations and restoration procedures. Transfer access through authorised methods without publishing credentials; copying code does not back up the database.
No. Call the operation directly using a synthetic account without the required permission and a known identifier. The server must enforce the rule independently of the interface. Keep the status and observed effect without publishing tokens, passwords or real data.
How can a repeated operation be tested?
Submit the same test operation twice under the agreed contract. Check stored data and side effects such as notifications, rather than the response code alone. Describe the rule distinguishing repetition, a legitimate new request and retry after failure.
What should a useful error response contain?
It identifies the operation, problem and next action without exposing sensitive data. Provide a correlation identifier if needed. Distinguish errors users can correct from incidents requiring technical intervention.