[WIP] MSC3956: Extensible Events - Encrypted Events#3956
Conversation
turt2live
added
proposal
| ## Proposal | ||
|
|
||
| **Author's note**: There is fairly strong, and reasonable, opposition to having encryption be a content | ||
| block. This could theoretically allow an event to be partially encrypted, which is undesirable. |
There was a problem hiding this comment.
I am not agreeing nor disagreeing with the sentiment here, but it is already possible to have partially encrypted events since m.relates_to is always in cleartext; it does create complexity though of how that interacts with a decrypted plaintext version.
There was a problem hiding this comment.
indeed, though the concern is more that someone will stick m.markup as a fallback on an encrypted event, which defeats the purpose of encryption. This proposal as-is also theoretically allows for weird parts of the event to be encrypted, though that's not intended to be valid.
|
|
||
| This allows the `m.encrypted` content block to be reused by other event types, if required. | ||
|
|
||
| For clarity, this is *not* intended to allow unencrypted fallback on an encrypted event - doing |
There was a problem hiding this comment.
I wonder if it would make sense to have the spec say that a server impl should refuse a cleartext fallback if m.encrypt is present.
While this doesn't protect against people modifying their servers this would protect against faulty client implementations.
There was a problem hiding this comment.
On further thinking and talking to people: wouldn't it make sense to even enforce this in the auth rules? Is there any case where checking wouldn't be trivial or where you need plaintext next to encrypted?
There was a problem hiding this comment.
Reasoning behind doing this via auth rules would be that this is a security concern. So this would prevent a security relevant issue which would justify having strict validation of the event imho as security outweighs most use cases I can think of.
There was a problem hiding this comment.
At this stage, auth rules feel overly strict for the general protocol: while I can agree with the principle that information should be enforced as encrypted, the general purpose nature of the protocol would call for a less constraining approach.
Also at this stage, I currently believe that adding auth rules to specific event types would be a design failure. Ideally, there aren't exceptions.
github-project-automation
Bot
moved this to Tracking for review
in Spec Core Team Workflow
| "sender_key": "<sender_curve25519_key>", | ||
| "device_id": "<sender_device_id>", |
There was a problem hiding this comment.
as i understand current spec, device_id is deprecated and if not already could be dropped by this MSC. sender_key is only used for olm (hence technically shouldn't be present in this example)
https://spec.matrix.org/v1.18/client-server-api/#mmegolmv1aes-sha2
5 participants
Rendered