Why a Radar feed URL is rejected: paths, queries and evidence-link differences
A clickable evidence URL is not automatically a valid Radar feed
CollieRadarFeed.validate permits a path but requires HTTPS, nonempty host, no embedded user/password, no query/fragment. The port may be omitted; if explicitly specified, it must be within 1–65535. Evidence uses separate parseSafeEvidenceURLs rules. Passing feed shape proves neither authentication, reachability, a successful read nor valid snapshot. No source service or credential setup is added.
Different from evidence admission and offline diagnosis
Evidence admission checks clickable candidates; offline review checks failures. Here whether a configured source address meets consumer boundaries comes first. No network request or real credentials are used, and shape is not private-source identity verification.
| Check | Verified scope | Manual review | Not established |
|---|---|---|---|
| Edge whitespace | Trim before parsing | Original configuration/returned URL | Every internal character cleansed |
| Scheme/host | HTTPS/nonempty host | Actual source identity | Trusted-domain allowlist or TLS test |
| Path | Allowed on feeds | Actual source endpoint | Browser-origin rules interchangeable |
| Credentials | No user/password | Existing authorized auth process | Passwords in URLs are supported |
| Query/fragment | Both absent | Even view=demo differs | Evidence allowance transfers to feeds |
| Port | Optional; if specified, 1…65535 | Access/reachability | Every port/source tested |
| Response | URL validation is not JSON decoding | Model/date/size/identity | Any web page is a snapshot |
| Failure | Original guards remain | Category/original configuration | Automatic HTTP/other service/private disclosure |
Synthetic address comparison
https://example.com/radar.json illustrates a path only. Adding ?view=demo or #part meets current query/fragment rejection conditions. Nothing is fetched and no real source is established.
Do not remove real authentication conditions or use an unapproved public service to pass a check. Shape, authentication, reading and content acceptance are separate; none silently supplies the others or write access.
FAQ
Can every clickable HTTPS link be a source?
No. Consumers differ, especially query/fragment rules.
Must a feed be a bare domain without path?
No. A path is allowed; actual endpoint validity still needs authorized evidence.
Is a harmless query allowed?
The current validator requires no query, irrespective of assumed harmlessness.
Does validate prove authentication?
No request occurs there. Identity, permissions and response remain separate.
Can failure be bypassed with HTTP?
Do not bypass HTTPS; inspect actual errors/authorized configuration.
Was the example source created or fetched?
No service registration, credentials or network request occurred.
Reading and source
Evidence URL admission · Offline/source failures · JSON consumer conditions · Start · Privacy
