X-Git-Url: http://jsfdemo.indexdata.com/?a=blobdiff_plain;f=NEWS;h=946f3f45d8df17cd0bd69e081e585ef631d863bb;hb=812bca889b5e5fb884b5a4785dce0d26095e099a;hp=2c40afcab2ea438d4d590a33a0a3dd243029f248;hpb=2d517dcaa085fcdcdbbf587b76d45a97eca35571;p=pazpar2-moved-to-github.git

diff --git a/NEWS b/NEWS
index 2c40afc..946f3f4 100644
--- a/NEWS
+++ b/NEWS
@@ -1,3 +1,80 @@
+--- 1.6.22 2012/10/11
+
+Requires YAZ-4.2.40 to support native solr support. 
+
+Fix and improve logic handling whether or not to re-do search on sort 
+order changes. A sort order with change in ascending/descending only 
+whould not trigger a new search, which is required for targets with 
+native sorting capabilities. Each client is now checked if instructions
+(sortmap) exist for native sorting and only client that does requires 
+it is researched. Other clients is just re-ingesting the records, 
+they already have. The resultset is now cleared if any researching 
+is done. 
+
+Connection sharing between session has broken since version 1.6.8 with 
+introduction of logic that would minimize searching if pazpar2 could 
+detect this based on same query and limits and partly sort order. 
+This could lead to segementations violations.
+
+Added a chapter in the manaul about relevance ranking.
+
+--- 1.6.21 2012/09/24
+
+Rank tweak: follow=number will increase mult by number if two terms
+occur next to each other; number-1 if they are one term apart , .. 0
+if they are number a part (all in order). Default is 0 (following
+terms has no effect).
+
+Rank tweak: lead=k will divide mult by 1 + log2(1+k*l) where k is
+value given by lead and l is length from beginning of field where
+term occurs (l=0 for first term, l=1 for second term, ..). Default
+value of k is 0.0.
+
+Rank tweak: length=strategy. length="linear" if mult is to be divided
+by length (existing, default behavior), length="log" if mult is to be 
+divided by log2(1+length), length="none" if mult is not to be affected
+by length.
+
+--- 1.6.20 2012/09/21
+
+Rank algorithm details may be printed as part of show response in
+element <relevance_info>.. This is only printed if <rank debug="yes"/>
+is used in service.
+
+Record as returned by show/record command have a minimal indentation
+which makes things human-readable.
+
+New configuration of default sorting criteria (sort-default) in 
+service definition. If no criteria is defined it will be as before
+'relevance'. 
+
+Search command now supports sort parameter just as the show command. 
+If no parameter is give, it will use the service sort-default value.
+
+--- 1.6.19 2012/09/18
+
+Rank algorithm skips strings that gets normalized to empty string.
+For example, & and ! could map to the empty string. The weight for
+those terms is now 0 (as if they were not part of the query).
+
+Rank algorithm does not use CCL from limitmap; only from the query
+parameter (user query).
+
+Obey -w dir.
+
+--- 1.6.18 2012/09/17
+
+Rank algorithm configurable by 'rank' element inside service. So far
+only, attribute 'cluster' is recognized. If cluster="yes", multiple
+records inside a cluster boosts higher than single records. This
+is default behavior and existing behavior. cluster="no" takes the
+average score of each record in a cluster.
+
+--- 1.6.17 2012/09/05
+
+Fix bad re-use of connections (connections with changing proxy should not
+be reused).
+
 --- 1.6.16 2012/08/22
 
 Fix a bug introduce in 1.6.15 around the position sorting. It resetted the