Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cf-dev] Release Plan for Californium ?

On Thu, 2016-11-17 at 15:46 +0100, Simon Bernard wrote:
> I know there is always this bug about blockwise transfer :/ I think this 
> should be fixed in the 2.0 release but we can do that without 
> refactoring all the stack.
> 
I think we should start to make a list of issues on GitHub which we think
need/should be in 2.0.0. I can create a "2.0.0" milestone in GitHub to which we
can then assign the issues.

However, we need to create specific issues for tightly scoped problems we want to
fix. Otherwise we will not be able to make any progress. As an example, I do not
think that an issue like "Fix Blockwise Transfer" will do any good. Instead we
should maybe start with "Add support for etags" and or "add blockwise transfer
state to ObservationStore" (if we believe that this is what we need to do).

We can then start to focus on the issues targeted for 2.0.0 and will (hopefully)
be able to do a 2.0.0 release in the near future.

> I have some concerns about this change, this is an attractive idea to 
> simplify the stack. But the more I think about it the more I have doubt. 
> Limiting the stack to a request/response handler which don't know 
> anything about blockwise and observe will not be so easy. E.g. when we 
> receive a response for an observe request we don't know if we must 
> return a RST or ACK if we don't know anything about observation.
> Anyway, this coapstack refactoring seems to be a really impacting task 
> with some risks as we don't really know where we go and if it will 
> really be better.
> We should maybe put this change on the  3.0.x branch ? and keep focus on 
> security and stabilized observation behind NAT for the 2.0 ?
> 
You are probably right. However, I think this mainly depends on whether we will
really be able to "fix" blockwise transfers in a clustered environment.
IMHO we need to add more state to the ObservationStore for this to work, e.g. the
 etag from the latest notification, the latest notification counter etc.

Kai

Back to the top