Research

This section holds work that is not settled. Everything else in the documentation states a position; these pages state a question, what has been looked at, and what is still unknown.

Keeping them separate is deliberate. A research direction presented as a recommendation gets adopted before it is understood, and then defended rather than evaluated.

Current investigations

A DSL for performance test writing

Can the workload model be separated from the execution mechanics, expressed in a language a non-specialist can read and review, and compiled to whichever runner executes it?

• Why a performance DSL — the problem, and the evidence that it is a real one rather than an aesthetic complaint.
• Survey of existing DSLs — what k6, Gatling, JMeter, Locust and Artillery already offer, and where each stops.
• A sketch — a concrete proposal, with the open questions it has not answered.

Campaigns as code

A campaign is a decision — scope, owner, entry and exit — and today that decision is made by clicking through Jira, where it cannot be reviewed as a diff. Can it be declared in the repository and applied to the test management tool?

• Campaign definition DSL — the declaration, and the excluded block that puts the risk decision next to the scope.
• CLI for the Xray API — what would apply it, and the four other operations every pipeline reimplements with curl.

These two are a pair and are pursued together: a declaration with nothing to apply it is a document, and an applier with nothing to apply is a wrapper. Both sit on top of Xray, which is what defines campaigns today.

Status

Investigation, both of them. No implementation, no commitment, no recommendation to adopt anything. The pages that describe what is actually done today — performance, performance tooling, management and reporting and campaigns — are unaffected by whatever these conclude.

Each investigation has the same three possible outcomes, in order of current plausibility:

1. A thin layer over what exists. The model in a declarative file, compiled to a k6 script, or applied to an Xray Test Plan. Smallest step, most of the benefit, and reversible.
2. A convention rather than a language. A documented structure with no new tooling at all. Free, and it may be enough.
3. Nothing. The problem is real but the cost of the answer exceeds it, and the conclusion is to write it down and stop.

Option 3 is a legitimate result and this section will say so if it is reached. A research section that never concludes “no” is a marketing section.

Both investigations also share a first step, and it is the same one: take four real artefacts — performance scripts, or the last four releases’ Test Plans — and express each by hand in the proposed form. What does not fit is the answer, and it is evidence rather than speculation.

Why it is published at all

Because the shape of an investigation is worth reviewing before its conclusion is. Someone reading the survey may know that one of these tools already solved this, which is a cheaper answer than the sketch — and they can only tell us if the work is visible.