Page MenuHomePhabricator

AbuseFilter should not let users specify arbitrary error codes which are then emitted over the API
Closed, DuplicatePublic

Description

The AbuseFilter lets users specify arbitrary strings which are then emitted as error codes over the API.

Example:

This is a headache for app developers and API consumers (like the mobile apps, and VisualEditor), as a random user can effectively make breaking API changes at a whim. It makes it hard to develop an app when there is no contract as to what kind of information the API can return.

The AbuseFilter shouldn't do this.