On Sep 14, 2005, at 5:31 AM, Tore Halset wrote:
> On Sep 13, 2005, at 22:16, Andrus Adamchik wrote:
>
>
>> We do need a neutral XML serialization mechanism (based on Kevin's
>> work, I assume). And with ability to do a seamless switch between
>> efficient binary serialization and slower but more flexible XML
>> serialization. Will be working on it...
>>
>
> Will some of the features only be availiable via the XML channel?
>
> - Tore.
Not Cayenne features per se. Serialization mechanism is outside of
the scope of the Cayenne client/server communication "protocol" that
only defines what messages server can process and what objects it
would send in reply.
But serialization mechanism choice would affect deployment. If we use
XML, client can be anything - Java, .NET, etc. Encoding can be done
in generic form... If we use Java serialization on the other hand (or
Hessian variety which is part custom, part Java), two communicating
JVMs would have to access the same set of classes, hence - client
classes would have to be available on the server.
Short of hacking Java binary serialization format, I can only think
of one workaround - encode objects as some sort of raw "data rows"
that use generic JDK classes and convert them to real objects on the
receiving end. I guess this would add more overhead, but still less
than XML.
Andrus
This archive was generated by hypermail 2.0.0 : Wed Sep 14 2005 - 18:58:55 EDT