[adamjohnwright]

Spec 015 (specs/015-graph-exports), T070.

What it does. /ContentService/exporter/event/{id}.biopax is answered from the library service's /biopax/{stId}, which T069 adds. It follows research R6:

• application/rdf+xml, downloaded as {stId}.biopax;
• SBML's 404 and 400 in the envelope, decided from the object's labels;
• the body streamed;
• its own share of two, apart from SBML's and the deep pool;
• a 503 when the service cannot answer, never a fallback.

How:

• SBML's answer is generalised into exportAnswer, with one format for each export.
• /exporter/event/{id} dispatches by extension: .biopax to BioPAX, .sbml as before, and anything else to Java as before.
• The service's address is still SBML_SERVICE, which now names the library service for both exports.
• parameters.ts's kind for the id now takes .biopax, so the answer is in standard form and cached per release (spec 011).

Not routed yet. T071 adds the route once T049 and T050 are recorded. Until T069 merges, the service has no /biopax/ and would answer it with a 404.

Tests: biopax-export.spec.ts, 10 tests, against a stub service. Each of six mutants made at least one fail:

• the two shares shared;
• an .owl download name;
• the service asked for SBML;
• the dispatch always going to SBML;
• a text/plain type;
• .biopax left out of standard form.

SBML's own 14 tests pass unchanged, and so do all 965 of content-node's.

Also, from #753's review (tasks.md):

• T032 now says it was next-release.sh cleanup that removed release 97.
• T072's 16 minutes cites its measurement.
• The dependencies say T072 needs a copy at beta's release.

🤖 Generated with Claude Code