My position is that when it’s open access, we might as well link the primary source. ResearchGate is generally legal. It’s the responsibility of the authors to upload accepted manuscripts if the final document is not open access. AFAIK it does not do anything dodgy like archive.is does.