Paper Number: C437
Title: The jCrunch(TM) Java Numerical Libraries
Author: William N. Reynolds

The paper describes an effort to produce high-quality numerical
libraries in Java.  The approach taken is to automatically translate
high-quality numerical libraries in Fortran into Java.  Up to three
versions of each function are created:
  (1) A pure Java version;
  (2) A JNI wrapper around a compiled Fortran version; and
  (3) A Native Broker version, which automatically selects one of the
      first two at runtime.

Recommendation: Reject.  While the existence of these libraries
  certainly merits a conference publication to alert the community, the
  paper is neither deep nor well-written enough to merit publication in
  a journal.  Furthermore, the most interesting questions raised by the
  paper are never addressed, such as the means by which the
  Fortran-to-Java translation was accomplished, and the treatment of
  related work is much too cursory.  The paper instead is almost
  entirely a description of the Java interface to the libraries,
  certainly not an exciting or particularly relevant topic.

Comments for the author
-----------------------
Major comments:

1. Related work is mentioned once, in the final paragraph of the
   Introduction.  However, it is not analyzed or described at all.  I
   don't know what was done in [1,2,3], nor what jCrunch provides that
   those works do not.

2. The paper begins by talking about translating Fortran to Java, in the 
   first paragraph of Section 2.  This is an important issue; it caught
   my attention and made me excited to read the paper.  But all I learn
   about this is that is has 2 drawbacks, and a few sentences apiece on
   how these drawbacks are addressed.  In my opinion, this topic should
   be explored in considerably more depth.  It is a much more important
   issue than the details of the Java interface to the libraries.

3. The sample application in the appendix is useless without some
   detailed commentary or explanation to go with it.  I don't know what
   I am supposed to learn by reading it.

Minor comments:

  Where is the abstract?

  Page 1, line 2: omit the apostrophe from "It's"

  Page 1, last sentence of section 1: "... four years after *the*
  introduction of Java ..."

  Page 1, paragraph 4, line 1: "The jCrunch libraries are *a* set of ..."

  Page 2, line 4: omit the apostrophe from "API's".

  Page 2: combine the last 3 paragraphs of Section 2.

  Page 3: the last example on the page is meaningless to me.  What is it 
  trying to demonstrate?  What do all those variable names mean?

  Page 4, last line: move to the next page.

  Page 6, line 7: "Subarrays" is missing the second 'a'.

  Page 6, second sentence of "Offsets and jCrunch": omit the
  parenthetical remark.  It will just confuse people.

  Page 6, next to last sentence: How am I going to note that?  This
  still looks like alphabet soup to me.

  Page 7, 1st sentence: Capitalize "every".

  Page 8, line 4 of "Drawbacks of Native Code": "segmentation
  violation", instead of "segmentations violation".

  Page 8, line 6 of "Drawbacks of Native Code": omit the apostrophe in
  "DLL's".

  Page 9, References: reference 3 should be a footnote instead.