- Looking for more contributors!
-
Biggest change since version -06 was removing the message status
format
- Split into a separate document. No objections from the room
-
6 open issues
- IANA registry of dispositions
-
Timo asked about MIMI Content Nesting
- Issue was closed because there was no activity
- Rohan suggests that there is a concrete proposal in the draft
that doesn't have obvious bugs.
-
Inconsistent timestamps (#60)
mimi-content uses uint4 in seconds, which will wrap in 2106
- Could do nothing, given we have a long time before it becomes an
issue
- Timo suggests that the issue is slightly different.
-
Message ID Calculation (#61)
- The method of concatenating variable length items could lead to
2 different messages leading to the same URI. Unlikely but
possible.
- Possible solutions include document the attack and tell hub and
client to validate
- Could also disallow
mimi:// in the path of a sender or room
URI
- Could add a separator which is not legal in URIs
- Could frame with CBOR
- Audience suggestion was to use CBOR
- Audience suggestion of a binary operation to make into valid
CBOR without the dependency
- Question from video: why include the hash algorithm into the ID
vs doing it out of band
- Suggestion to include hash algorithm in versioning rather than
the ID
- Rohan makes the case that chat history should be self-describing
and can't be negotiated
- Back and forth around whether there is somewhere else we could
store the message_id algorithm
- Konrad points out that the application could store it alongside
the message_id
- Konrad also suggests including the version in MIMI Content
- Rohan wants the original 7 fields to stay fixed
- Back and forth about whether the hash algorithm would be an
extension or part of the core algorithm
- Audience suggestion of using the OID
- Audience question about whether an attacker would have the same
ability to tamper with the field in another place or in the ID
- Question about whether it would be a reasonable thing to
recalculate the ID on messages you receive
- Discussion about where to place the field and if the proposed
solution is in line with Deirde's suggestion
- Rohan and Deirdre agree the best place is in the extensions
-
Replying to Edited messages (#67)
- What message ID do you put in the
InReplyTo for a reply or
reaction when the message has been edited?
- Rohan's suggestion is the one that was replied/reacted to
- Marvin says that they have previously used the first message in
other implementations
-
Once the message_id issue is resolved and the other issues are
closed, it should be time for a last call
## Nested Content {#nested-content}
Presented by Timo
-
Showed an example of a simple image and caption
- There is some desire to have an internal references
- References make things more powerful
- No references gives the application more control. Internal
references give the sender more control
- References can lead to strange situations like self-references or
references to invisible components like an un-selected chooseOne
- Marvin suggests that internal references should only be used for
fancy cases. Simple cases should not use references
- Mojtaba suggests that verification would be hard but might be worth
the trouble
-
Timo's proposal is to have an attachments section that can be used
as a reference. Not everything could be referenced
- Attachments would not be able to reference each other
- Would not have a fallback for clients without markdown support
- Requires understanding of Markdown
-
Rohan says that there are many real-world use-cases for references
as described in previous discussions.
- Bigger question about whether we want a bigger renderer that can
handle things like polls, or whether that should be handled by the
application
- Alissa: Closing down the flexibility of the existing spec would need
a clear reason.
- Rohan suggests that the design in the doc today is implementable and
tractable.
- Risk to interoperability with a very generic approach. Applications
would need to support everything to fully support the content type.
- Timo suggests that there is a very expressive syntax but lacks a
grammar of how to use it
- Rohan: in order to get interoperability you need to be able to
render the content you receive. Being able to restrict things with
unreasonable semantics is required. Content negotiation would handle
cases of unsupported media types.
- Timo would like to have more official examples of things like polls.
Maybe not enough interest to have it fully specified in MIMI itself.
- Questions about the
documentChooseOne and whether it is valid
markdown
- Alissa suggests putting together a concrete proposal of an alternate
algorithm instead of examples
- Ambiguity about whether or not rendering Markdown as plaintext is
acceptable: Rohan clarifies that it is fine to render to plaintext.
- Raphael suggests that it might be premature to move content to last
call. Don't have interop yet and other issues might come up.