💾 Archived View for rawtext.club › ~sloum › geminilist › 001278.gmi captured on 2020-10-31 at 02:10:18. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2020-09-24)
-=-=-=-=-=-=-
Martin Bays mbays at sdf.org
Tue Jun 2 23:33:13 BST 2020
- - - - - - - - - - - - - - - - - - -
At this point I am honestly considering speccing that =
lines may
*only* use URLs whose scheme corresponds to an application protocol.
If the result of limiting schemes is that clients just ignore lines with other schemes, you have a clear opportunity for drift. Some clients could start to go against spec and interpret links which don't fit this restriction.
Perhaps instead, links with whacky schema shouldn't be disallowed, but there should be clear requirements for how an interactive client interprets them. I suggest two rules:
1. NO MAGIC: an interactive client must do nothing with a link URI other thandisplaying the text of the URI, until the user expressly follows the link.
2. The URI of each link line must be displayed in full.
The first rule makes links with arbitrary schemes relatively unobjectionable. The second rule is to stop drift away from the first rule, by cutting off what could otherwise be a channel for communication between off-spec clients and servers which doesn't affect on-spec ones.