Re: Byte and char types

On Fri, 2006-06-16 at 16:29 -0700, Thomas Wang wrote:

> Well, I personally do not like any of the new constructs. For practical
> reasons, Hessian has not seen wide enough adoption to be adding
> extensions to it. That will hinder adoption.

The proposed changes should only help adoption. All Hessian 1.0 clients
and servers will work with Hessian 2.0. So even if the only Hessian 2.0
implementation is our Java one, everyone will still play together
nicely.

There are no semantic changes planned. In particular, the encoding is
entirely invisible to the user. The only changes are for those protocol
implementers who want to add a more compressed version.

By the way, a draft of the grammar is at
http://wiki.caucho.com/Hessian_2.0_Grammar

There's also some data comparing Hessian 2.0 to Java serialization at
http://forum.caucho.com/node/52

The next Resin snapshot will have a Hessian 2 implementation (probably
Friday)

> The CESU-8 encoding used by Hessian is still somewhat a problem for
> cross-platform support. You can have the first part of a 16 bit
> surrogate in the end of a chunk, and the second part of the surrogate
> in the beginning of the next chunk. You may not be able to get
> an array of unicodes just by processing at the chunk level.

I've added this as a bug report. We can add the requirement that
surrogate pairs must be in the same chunk.
http://bugs.caucho.com/view.php?id=1203

-- Scott

>
> Thomas Wang
>
>
>
Received on Mon 19 Jun 2006 12:21:21 -0700