2.0.8.0
minorVersion 2.0.8.0 makes the domain reputation check query the domain's own registry directly, and fixes Deep Scan spinning during the model download.
What changed
- The registration-date check now asks the domain's own registry directly, not a shared redirector
- One fewer third party in the lookup path: only the registry serving the domain's ending is contacted
- Deep Scan now reports download progress to a scan that starts while the model is already downloading
- Deep Scan no longer spins forever with no result if you click it during the first download
- Domain endings with no published registry - .ru and .рф among them - can't be dated, and are never flagged for being new
The registration-date check is what catches a throwaway domain on an ordinary-looking address: a scam that would sail past a blocklist of suspicious domain endings. It works by asking the domain registry when a domain was created - which means it is only as dependable as the thing it asks.
It used to ask through a single shared community redirector, and that was the source of the flakiness. Under a burst of lookups the redirector would start refusing them; it turned away anything that didn’t look like a browser; and for whole sets of domain endings it had no answer at all, which meant the check quietly returned nothing rather than failing. Now the extension looks up which registry serves the domain’s ending in IANA’s public list of registries, and asks that registry directly. One fewer party in the middle, and no redirector to be throttled by.
That also makes the boundary honest and worth stating: some domain endings publish no registry to ask.
.ru and .рф are among them - the .ru/.рф registry serves registration lookups to registrars
only. A domain whose age can’t be determined is never flagged for being new, so this is a missing
warning rather than a wrong one.
This release also fixes Deep Scan on the one path where it could hang. The model downloads once, about 67 MB on first use, and a scan clicked while that download was still running used to join it with nowhere for the progress to go: the spinner ran, no result ever arrived, and nothing was logged. Those scans now receive the download progress as it happens and finish normally.