IBM InfoSphere Global Name Recognition
GNR 4.1.0.1 GA Fixpack
December 2009


ipsdb00060641
APAR # PJ36465
APAR # PJ36467
ipsdb00061077
ipsdb00061063
ipsdb00061222
ipsdb00060431
ipsdb00060457
ipsdb00057426
ipsdb00057332
ipsdb00060462
ipsdb00056774
ipsdb00061426
ipsdb00061104
ipsdb00059864
ipsdb00061399


This readme includes the following sections:

	   I. Notes
	  II. Required files
	 III. Issues corrected in this fixpack
	  IV. Installing the fixpack

==============================================================

I. Notes:

This fixpack was built for all supported platforms and applies to the following IBM GNR products:

 - IBM InfoSphere Global Name Manager version 4.1
 - IBM InfoSphere Global Name Analytics version 4.1
 - IBM InfoSphere Global Name Scoring version 4.1

You must have v4.1 of one of these products installed prior to installing this fixpack.

In addition to new fixes, this fixpack contains GNR 4.1.0.0_hotfix1 and includes all fixes since the release of v4.1 in April, 2009. 


==============================================================

II. Required files:

Use the fixpack installation media to install this update.


==============================================================

III. Issues fixed since 4.1.0 GA:

1. Issues that are corrected in this fixpack (4.1.0.1):

a- [ ipsdb00056774 ] NameWorks now treats names categorized as Unknown as personal names instead of organizational names.

b- [ ipsdb00057332 ] NameHunter distributed search no longer reports all matches as coming from regularized alternate parses in original name data.

c- [ ipsdb00057426 ] NameWorks now defaults compressedBitSig to true as intended.

d- [ ipsdb00059864 ] NameSifter: necessary changes were made to Sifter Rules file.

e- [ ipsdb00060431 ] A more informative error message is provided when ds.config file uses the same name for two searchers (e.g., [searcher1] and [searcher1]).

f- [ ipsdb00060457 ] NameWorks now supports disabling categorization if not needed (such as when you only search personal names).

g- [ ipsdb00060462 - PMR 61906,019,866 ] NPP no longer misses relevant alternate parses.

h- [ ipsdb00061104 ] NameSifter: Correction/editing of values in ON-only list.

i- [ ipsdb00061399 ] NameHunter Server core-dump issue on AIX now resolved.

j- [ ipsdb00061426 ] Restored lost Chinese variants

 - - - - - - -

2. Issues that were corrected in hot fix 4.1.0.0_hotfix1):

a- [ ipsdb00060641 ] NVG uses erroneous Chinese variants in NH's vardata tables
NVG does not use the variant pair score values. The result is that pairs like CHEN / PI, which are totally unrelated, are returned by NVG as variant forms of each other.

b- [ APAR # PJ36465 ] Can't have too many NameSifter objects in memory simultaneously
Every time we create a new instance of NameSifter, it maps the NDO file into a new memory region even if it's already mapped into memory.  After some small number of times, we run out of address space for memory mapped files and die. 


c- [ APAR # PJ36467 ] Excessive memory growth in multi-threaded applications
In a multi-threaded application using NameHunter you will get excessive memory growth.
The v3.x-compatible wrapper for v4.1 also does not return culture information in the 'SearchResult' objects.

Summary - This situation:
* Only happens with multi-threaded applications
* Eventually stabilizes
* Memory table corruption, not leaked memory


d- [ ipsdb00061077 ] The 'DecisionVector' used to hold return values is not cleared before new culture codes are added
If a client application reuses a 'DecisionVector' object, multiple 'decide()' calls append new culture codes to the existing list.


e- [ ipsdb00061063 ] Latin transliterator module lets through things outside the basic 26	
   NameTransliterator's Latin transliterator module isn't correctly filtering characters. The CharsHandled and CharsTolerated sets should only be letting through the characters in Unicode's ASCII, Latin-1, and Latin Extended A blocks, plus a smattering of characters from the other blocks that have canonical or compatibility decompositions down to the 26 letters in the basic English alphabet. The current rule set is also letting various characters in the IPA and Spacing Modifier Letters blocks.
   
f- [ ipsdb00061222 ] Added in ICU libraries for transliteration.


==============================================================

IV. Installing the fixpack

This fixpack is installed by following steps similar to the base product 
installation.  

1. Obtain the IBM InfoSphere Global Name Relationship product media. 

2. Run the install program: 
   
   GUI option: 

   a) In GUI mode from a command line Navigate to the /Disk1/InstData/VM/ directory 
   on the product media. 

   b) Run the install executable. 
 

   Command line option: 

   a) Navigate to the /Disk1/InstData/VM/ directory on the product media.

   b) Run the install executable with the -i console option. 
 

3. Follow the instructions on the install program wizard.

   When prompted, indicate the path/directory of your v4.1 installation.