On 8/6/2013 8:37 AM, Guido Medina wrote:
I saw inside the solr.war file there is a protobuf version 2.4.0a, I
have two questions about it:
1. Where does Solr uses protobuf? And is it better than HTTP?
2. Why is it such an old version if protobuf recommended versions are
2.4.1 and 2.5.0 - 2.5.0 has an extra 25% performance over 2.4.1
according to release notes:
http://protobuf.googlecode.com/svn/trunk/CHANGES.txt
A quick 'grep -rl' on the lucene/solr source code seems to indicate that
protobuf is a dependency of hadoop. Hadoop 2.0.5 is included in Solr
because version 4.4 can store indexes in HDFS instead of on a standard
filesystem.
It appears that hadoop has problems with the newest protobuf:
https://issues.apache.org/jira/browse/HADOOP-9346
Thanks,
Shawn