My plan was to, as you say, "literally re-implement Metakit in Java". I'm not interested in a solution that is not in pure Java.

As to your qualms.
1) I agree that keeping the features in step with Metakit would require a significant effort. On the other hand, at the moment I am not using nearly all the features of metakit and would rather have a partial implementation than no implementation. Also, the progress I've made in a few hours suggests that a basic implementation would not be a huge amount of work. For example, using java.nio ByteBuffer and FileChannel.map has allowed a very quick implementation of memory mapped metakit files. Finally, I think metakit is cool enough that quite a few Java people might be interested in helping.

2) In some applications Java performance is really very good (and often underestimated). And the fact that a great deal of use of Metakit is made through scripting interfaces that are slower than the native C++ versions suggests that a native Java version would be quite usable for many applications.

On the other hand, I am discussing with jcw some alternative methods for getting a Metakit type solution in pure Java.

cheers,

Bruce

On May 12, 2005, at 10:44 AM, [EMAIL PROTECTED] wrote:

I'm a little unclear about the project scope here. �Do you intend to literally re-implement Metakit in Java? �Or do you intend to write an effective and easily used Java "wrapper" for the underlying Metakit C++ library (like the Python wrapper)?

If the latter, I can see the advantage. �If the former, I'm a bit skeptical about the result. �There are two qualms I have with it. �The first is that this would require substantial effort to keep its features in step with evolving Metakit features. �The second is performance.

<unknown.jpg><mime-attachment>_____________________________________________
Metakit mailing list - [email protected]
http://www.equi4.com/mailman/listinfo/metakit

Bruce A. Johnson, President
One Moon Scientific, Inc.
839 Grant Ave.
Westfield, NJ 07090

Phone 908 517-5105
Fax 908 517-5107
Email [EMAIL PROTECTED]
Web www.onemoonscientific.com
_____________________________________________
Metakit mailing list  -  [email protected]
http://www.equi4.com/mailman/listinfo/metakit

Reply via email to