if by "old behavior" you mean incremending the start param then the the order of results when doing concurrent indexing was always dependent on what exactly your sort was.
when using a cursor, the impacts of concurrent indexing are also dependent on what your sort clause looks like -- but in differnet ways. both situations are extensively documented... https://cwiki.apache.org/confluence/display/solr/Pagination+of+Results#PaginationofResults-HowBasicPaginationisAffectedbyIndexUpdates https://cwiki.apache.org/confluence/display/solr/Pagination+of+Results#PaginationofResults-HowcursorsareAffectedbyIndexUpdates : I have a quick question about this new implementation - in the old : implementation AFAIK, in a real-time indexing scenario, the results : gathered from paging would not be consecutive. Meaning you would ask for : 50 docs, new docs arrive, when you ask for the next 50 docs - you get an : arbitrary new document set (50 after any newly inserted docs). : : Having read a bit about the cursor implementation, is it true that the : next 50 results are now consecutive to the first set due to the fact : that lucene actually tracks the mark? -Hoss http://www.lucidworks.com/
