You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
data.isamples.org currently serves parquet files flat at the root (e.g., data.isamples.org/isamples_202601_wide.parquet). Ben Norton suggests adding path segments that convey resource type, following OGC-style patterns:
data.isamples.org/parquet/isamples_202601_wide.parquet # data files
data.isamples.org/record/<uuid> # individual sample records
data.isamples.org/term/<term-slug> # vocabulary terms
"This allows you to better manage resources, provides additional context and informs the user what type of resource a pid is expected to return. This pattern is also part of several specifications (i.e. OGC)."
— Ben Norton
Context
data.isamples.orgcurrently serves parquet files flat at the root (e.g.,data.isamples.org/isamples_202601_wide.parquet). Ben Norton suggests adding path segments that convey resource type, following OGC-style patterns:Relevant specifications
/collections/{id}/items/{recordId}Questions to discuss
Scope vs. timeline — The grant ends July 2026. Which path segments are realistic to implement?
/parquet/for data files: trivial (Worker routing change + redirects from old flat paths)/term/for vocabulary: moderate (could redirect to existing vocab pages on the site)/record/<uuid>for individual samples: heavy (requires a query service, not just static files)Backwards compatibility — PR Use data.isamples.org for all parquet file URLs #79 just migrated all references to flat URLs. If we restructure, we'd want redirects from the old paths.
Content negotiation — Should
/record/<uuid>return JSON-LD vs HTML based on Accept header? That's the full linked-data pattern but adds complexity.Versioning — Current files are date-stamped (
202601). Should the URI structure make versioning explicit (e.g.,/parquet/v202601/wide.parquet)?Current state