Re: [ecf-dev] Datashare channels...

[Scott Lewis <slewis@xxxxxxxxxxxxxxxxx>]

Hi Remy,

Remy Chi Jian Suen wrote:
> Hi Scott,
>
> On Mon, Apr 20, 2009 at 4:43 PM, Scott Lewis <slewis@xxxxxxxxxxxxxxxxx> wrote:
>   
> > Because the XMPPID serialization/deserialization is done by the XMPP
> > provider...and those classes (XMPPID) are available in the class namespace
> > of that provider.
> >     
>
> Sorry, can you put that in layman's terms? Did you mean that the
> message has already been deserialized once in the SOContainer?
>   

No...actually I was thinking of the shared object API (rather than the 
datashare API) when I answered this question...and in the shared object 
API the serialization/deserialization is done within the container (and 
this allows messages of the form:  sendMessage(ID target, Object 
message).  I've recently added more flexibility in the shared object 
container incidently, to allow for custom serialization within the 
container:  https://bugs.eclipse.org/bugs/show_bug.cgi?id=272224

But with the datashare API the payload of a send/receiveMessage is a 
byte array...i.e. IChannel.sendMessage(ID target, byte[] message)...so 
clients/users of the datashare API do any object 
serialization/deserialization on their own...and so they have to define 
their own available classes (via Eclipse-BuddyPolicy: global...as you 
discovered for docshare...which is a client of the datashare API).

> I also noticed that I was wrong and that both the XMPP provider and
> the DocShare bundles have a 'Eclipse-BuddyPolicy: global' in their
> MANIFEST.MF files. I added that in bundle B and now everything is
> working. :O I didn't have to put it in the datashare implementation
> bundle though.
>
>   
> > As far as I'm concerned, it would be useful to have more than one nio-based
> > provider/implementation.  Where is the nio implementation that you've been
> > working on (Remy)?
> >     
>
> At the moment, it's just on my hard drive and it's very much tied to
> the provider I'm currently working on enhancing.
>   

Ok.  When you can, it would be good to get it into the repo.

Thanks,

Scott