I don’t think we should start from a blank page; we could always create a
whole new project todo that.  I think we will be doing our jobs If we can
focus on improving pain points and stability.

I don’t have a problem with SSL being a filter; In fact, it is the most
flexible way to do it; just needs to be designed better.  Not only is it
overly complicated in implementation, there are several pieces of code
which scream “this will probably cause a concurrent memory problem”.

On Fri, Jun 22, 2018 at 8:54 AM, Emmanuel Lécharny <[email protected]>
wrote:

>
>
> Le 22/06/2018 à 14:01, Jonathan Valliere a écrit :
> > At what point is the SSL going to be redesigned and anyone using it will
> be
> > forced to update their code?
>
> Sadly, we have thousands of peope using MINA as it is.
>
> Mina 3.0 was an effort we started a few years ago to redesign this piece
> of code.
>
> >
> > Are we going to continue to support other projects which use internal
> APIs
> > or variables?
>
> I don't think any projects are using MINA internals. The change I
> reverted was a mistake I did. The local variable name made it easy to
> shot you in the foot, though.
>
>  Seems like every good idea Emmanuel had was reverted for
> > reasons from “using a private event you shouldn’t” to “not wanting to add
> > an empty function handler”.  Is there some official policy on
> compatibility?
>
> The policy is : "do not break the API in 'fix' versions", aka the Z in
> MINA-x.y.z
>
> This is why I forked a 2.1.0. We can play in this branch as much as we
> like.
>
> Although I think it would be better to keep Y versions for added
> features, still keeping the API ascendant compatible, and use a X update
> for breaking the API.
>
> I do think we should do what you suggested :
> - rename the 3.0 branch to something related to a coming future version
> - really start from a blank page (which is what we did with the MINA 3.0
> effort, btw).
>
> FTR, MINA 3.0 was expected to keep the idea of filters, but the SSL part
> was not designed as a filter. Also we wanted to have MINA handling
> either non-blocking IO and also blocking IO, to make it a generic
> framework, instead of being a NIO framework.
>
> The hard part are :
> - obviously find people interested in being part of the effort in the
> long term
> - keeping some kind of the existing MINA logic
> - simplifying the existing design which is, IMHO, way too complex
> - remove as much contention as we can
>
>
> That is a lot of work...
>
> --
> Emmanuel Lecharny
>
> Symas.com
> directory.apache.org
>
>

Reply via email to