-
Notifications
You must be signed in to change notification settings - Fork 3
⚗️Review BAM/COMIT protocols post RFC1-3 #23
Comments
To be moved to comit-network/RFCs once repo is open sourced |
I'll put this one on hold, so here are my current thoughts: The main design flaw seems to be that there is inheritance in place:
I am very confident that if we can figure out, how we can do composition instead of inheritance on a design level, this problem described in the beginning of this issue will be resolved. Another way to put this is that the concepts defined in one RFC (like BAM) spawn multiple design documents. They define abstract concepts that are later specialized in other RFCs. If we were to define the protocol in a way that leaves the concepts within one document, the design would be cleaner which would probably solve the above mentioned problem. I'll organize a discussion meeting to further dive into this. |
@LLFourn and I had some very productive discussions about this yesterday. For historical purposes, I am posting the image here. I'll submit a PR to the relevant RFCs as a follow-up of this that implements those changes. |
Now that we have specified two RFCs which use BAM it would be good to investigate how well it solved the problems it was designed to and if are we using it in the right way.
A specific topic that must be addressed the design of and our use of status codes, particularly with regards to the "reason" headers. We had a discussion of this and there were multiple ideas that are simply here as a reminder.
timeouts-too-tight
. We could partition them according to the "layer" they occurred to create some separation.NF**
"negotiation failed" to indicate the proposed action was understood but was rejected.Note this ticket isn't meant to imply that there is something wrong with the way things are now.
DoD:
Blocked by #7
The text was updated successfully, but these errors were encountered: