dev@fi.java.net

Re: XTech 2005: XML, the Web and beyond paper

From: Paul Sandoz <Paul.Sandoz_at_Sun.COM>
Date: Thu, 04 Aug 2005 10:37:44 +0200

Hi Thomas,

[CC'ing Ias and the FI dev email list]

As coincidence would have it Ias (Changshin Lee) has just committed a
fork of the J2SE fast infoset (FISE) StAX implementation to the
workspace. This is called FIME and it is planned to modify the code to
be more suitable for J2ME.

Ias presented his work at ApacheCon a couple of weeks back. He showed a
Web service using FIME on a mobile phone.

There are a whole bunch of stuff in the FISE implementation that is
probably not required in the FIME implementation, hence rather than
initially making FISE and FIME share a common code base we thought it
would be best to fork. Then afterwards we can take a step back and look
at what common aspects are being shared to see if it is worthwhile
merging a subset of code.

I suggest you add yourself to the FI project as an Observer initially.
Then subscribe to the dev_at_fi.dev.java.net email list [1].

You are right there has not been much traffic on the forum, of late it
there has been posts not related to Binary XML. I will also reply to
your forum post for the record :-)

Paul.

[1] https://fi.dev.java.net/servlets/ProjectMailingListList

Thomas Skjølberg wrote:
> Hi,
>
> I've read your XTech paper and downloaded your Fast Infoset source code
> and it all looks very promising. A short introduction:
>
> I am a Ph.d. student at the Norwegian university NTNU (www.ntnu.no)
> institute for signal proccessing (www.tele.ntnu.no). In addition, I am
> involved with a small startup company called Adactus (www.adactus.no -
> the web pages are all old now) which is into MPEG-21 - a standard
> defined in XML. My skills include low level xml parsing and info-theory.
>
> You're reading this mail for two reasons:
>
> - I'd like to try out Fast Infoset on a MPEG-21 client which I am
> developing, mainly for the J2ME platform. We will be testing on mobiles
> as well as benchmarking/evaluating a 3rd party chip pretty heavily.
> Our current version uses a custom kXML parser as well as custom DOM
> nodes. However, your implementation uses classes that are not defined
> in the J2ME CLDC (the J2ME java.lang equivalent) and thus do not
> compile. I think it will be straightforward to fix this but I am not
> sure (understanding that J2ME compatiblity should not kill performance
> on the J2EE / J2SE platforms). Two questions:
>
> 1. Have you even considering making the StAX code part J2ME compatible?
> Only a few interfaces are required in JSR-173 (although it may be more
> work than it seems).
> 2. To speed up things, could I help make those changes? Naturally, you
> don't want unknown people modifying your cvs, so I'd start by
> identifying the incompatible parts, propose solutions etc.
>
> - The MPEG-21 workgroup is looking into its own binary/streaming format.
> I think technologies like FI makes that a (really) bad idea and I need
> the means to prove my arguments and perhaps discuss certain techincal
> issues that may arise. The FI on java.net does not have its own forum
> and the xml binarization forum mentioned on the project home page looks
> deserted.
>
> Have a (continued) nice summer!
>
> Yours,
>
> Thomas Skjølberg
> Adactus / IET

-- 
| ? + ? = To question
----------------\
    Paul Sandoz
         x38109
+33-4-76188109