[
https://issues.apache.org/jira/browse/LUCENE-9116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17013007#comment-17013007
]
Adrien Grand commented on LUCENE-9116:
--------------------------------------
David, I'm worried that we might be setting a precedent here. Today we are very
permissive when it comes to adding a new postings format to Lucene, like
UniformSplit recently. I like it this way, it helps drive innovation, and
hopefully some of the ideas of these experimental formats eventually get merged
into the default codec like the pulsing optimization. I don't think we should
make experimental formats harder to remove than add, otherwise they get in the
way of improving the default codec, which is wrong. I called out that I was
removing these formats in a comment, I can send a notice to the dev list next
time to give it more visibility.
On a separate note, I think the Solr docs should be explicit that non-default
codecs/formats are not supported backward-compatibility-wise. Users might be
surprised otherwise to get corruption errors when upgrading to a new minor?
> Simplify postings API by removing long[] metadata
> -------------------------------------------------
>
> Key: LUCENE-9116
> URL: https://issues.apache.org/jira/browse/LUCENE-9116
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Adrien Grand
> Priority: Minor
> Time Spent: 40m
> Remaining Estimate: 0h
>
> The postings API allows to store metadata about a term either in a long[] or
> in a byte[]. This is unnecessary as all information could be encoded in the
> byte[], which is what most codecs do in practice.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
