On 2/15/07, Otis Gospodnetic <[EMAIL PROTECTED]> wrote:
One way around this is to get support for ParallelReader (I believe
ParallelWriter is still in JIRA, contributed by Chuck) into Solr.
http://lucene.apache.org/java/docs/api/org/apache/lucene/index/ParallelReader.html
Then you'd keep your big fields in one index, and the frequently modified and
shorter fields in another index. But I never understood how you'd keep doc IDs
in sync between the two, which is something that ParallelReader requires.
Aye, that's the rub.
ParallelReader keeps popping into my head too, but then I think about
what it takes to keep those id's in sync, and it seems like everything
needs to be re-indexed in the smaller index on a change to that index.
It doesn't seem easy or fast/scalable. I'd love to know what Chuck
is doing with this stuff.
-Yonik
