I wasn't thinking about making MINA 2 Codfec stuff comatible with MINA 3
ones, just wondering how we share the framework independent  API.
If you want to use the codec, with.. let's say plain java.net.Socket
blocking stuff. It would be stupid to depend the whole MINA3 core ?
Julien


On Mon, Jan 14, 2013 at 11:06 AM, Emmanuel Lécharny <[email protected]>wrote:

> Le 1/14/13 10:55 AM, Julien Vermillard a écrit :
> > I started refactoring MINA 3 HTTP Codec, but I wonder how to share the
> > codec infertaces (ProtocolDecoder.class and ProtocolEncoder.class)
> between
> > MINA 3 and MINA 2 ? a separate module ?
> > Because if we don't do that we will finish by having the two version
> > depending of each other like with the current LDAP codec.
>
> The issue with MINA 2 interfaces is that it requires additional parameters
> :
>
>     void decode(IoSession session, IoBuffer in, ProtocolDecoderOutput
> out) throws Exception;
> and
>     void encode(IoSession session, Object message, ProtocolEncoderOutput
> out) throws Exception;
>
> The ProtocolDecoderOutput/ProtocolEncoderOutput are callbacks.
>
> I would suggest we use simpler methods :
>
>     Object[] decode(IoSession session, IoBuffer in) throws Exception;
>     Object encode(IoSession session, Object message) throws Exception;
>
> where the returned Object whould be the decoded messages for the decode
> method, and the encoded message for the encode method.
>
> Woudl we have more than one message being decoded, we will get an array
> of decoded messages.
>
> I do think it's way easier to use than using a callback...
>
>
> Regarding the compatibility with MINA 2, forget about it. I have
> migrated ApacheDS from MINA 1 to MINA 2 3 years ago, and despite some
> kind of attempt to make MINA 2 quite API compatible with MINA 1, it
> required a full rewrite. At some point, you'd better accept the idea to
> review your code and switch. ALl in all, it took me a couple of days,
> max. I must say that switching to MINA 3 does not frighten me *at all*.
>
> Let's not make MINA the next Windows, still assuming some kind of
> inheritence with MS/DOS 2.1...
>
> --
> Regards,
> Cordialement,
> Emmanuel Lécharny
> www.iktek.com
>
>

Reply via email to