- Python 84.8%
- HTML 14.7%
- Mako 0.5%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
| docs | ||
| src | ||
| .gitignore | ||
| COPYING | ||
| COPYING.LESSER | ||
| MANIFEST.in | ||
| pyproject.toml | ||
| README.md | ||
splent_feature_curation
The review queue of an OpenHubs hub (GV-2): what a curator looks at before the hub acts on it.
- Publications. Where dataset's
DATASET_PUBLISH_POLICYiscurated, an account curation does not trust sends its dataset to review instead of publishing it. It waits, submitted, and nothing reaches the provider until a curator approves it; it is then published in the background, through the provider its owner chose. A rejection takes it back to a draft its owner changes and sends again, and its owner can take it back from review too. A dataset waiting for review cannot be changed: what is approved is what is published. - Withdrawals. The owner of a published dataset asks for it to be
withdrawn from its page (the button fills dataset's
dataset.view.actionsslot), saying why. A curator withdraws it (its DOI then answers with a tombstone that says why, 451 for a legal reason) or rejects the request, and the dataset stays published. - The queue, under Administration in the panel, for the accounts whose
role is in
CURATION_REVIEWER_ROLES, with the count of what waits on its menu entry. Every decision carries a note; the owner is mailed it, and the reviewers are mailed when something new waits (with the mail feature).
curation is the reviewer of dataset's policy: it implements dataset's
PublicationReviewer and registers itself from init_feature with
register_publication_reviewer, so a product whose policy is curated
installs it, or it does not start. It changes datasets only through
dataset's services, and keeps the requests and who decided them in its own
table, curation_request. It requires auth and dataset 0.8.0 or later, and
mails with the mail feature when the product has it.
| Route | |
|---|---|
GET /admin/curation |
The queue |
GET /admin/curation/<request_id> |
One request, with the dataset it is about |
POST /admin/curation/<request_id>/approve, .../reject |
The curator's decision, with its note |
POST /datasets/<dataset_id>/withdrawal-request |
The owner asks for a withdrawal |
POST /datasets/<dataset_id>/withdrawal-request/cancel |
The owner takes the request back |
Who is trusted
CURATION_TRUSTED_ACCOUNTS says which accounts publish without review on a
curated hub:
| Value | Trusted |
|---|---|
none |
no account: every publication is reviewed |
published (default) |
an account with a dataset published on the hub |
carried_over |
an account carried over from an older site (auth's account_origin) with a dataset published, the history uvlhub 2.12 gave it (D4) |
Whatever the value, a curator is trusted, and an account one of whose
datasets was withdrawn as spam or malware is not. A trusted account still
confirms its address first, as dataset's verified asks.
Parameters
All three are typed in [tool.splent.config_schema] and of derivation
bind: the product decides them, and an environment that changes them is
drift.
| Key | Type | Default | |
|---|---|---|---|
CURATION_TRUSTED_ACCOUNTS |
enum: none, published, carried_over |
published |
who publishes without review |
CURATION_REVIEWER_ROLES |
list of admin, staff |
["admin"] |
the roles of the curators |
CURATION_NOTIFY_REVIEWERS |
bool | true |
whether the curators are mailed what is new. The owner is mailed every decision either way, with mail |
In the model
An optional curation node of repositorio in openhubs_spl, which
publish_curated (the curated value of DATASET_PUBLISH_POLICY)
requires: publish_curated => curation, so a curated hub without its
reviewer does not validate. docs/openhubs_spl.uvl.patch was the change of
splent_spl_openhubs/openhubs_spl.uvl that added it.
Develop
splent feature:test curation --with dataset
splent feature:contract splent_feature_curation --write
It needs dataset 0.8.0, the release with the policy: until it is out, its
tests take dataset's working copy (--with dataset).