[
https://issues.apache.org/jira/browse/HBASE-859?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12629841#action_12629841
]
stack commented on HBASE-859:
-----------------------------
After the HBASE-868 and HBASE-877 escapades, we might need a METAHSK and
possibly a ROOTHSK. While a normal HSK has row/column/ts, a METAHSK would have
table/row/column/ts or maybe even table/table-row/table-ts/column/ts; in other
words, rather than rely on breaking row on delimiter, instead, rely on field
lengths. The new classes would know how to compare themselves. The two types
should never be mixed since their detail is contained beneath a Store (I
suppose, clients would have to know how to deal w/ the new type).
> HStoreKey needs a reworking
> ---------------------------
>
> Key: HBASE-859
> URL: https://issues.apache.org/jira/browse/HBASE-859
> Project: Hadoop HBase
> Issue Type: Improvement
> Reporter: stack
> Priority: Critical
>
> HBASE-832 had to make an ugly hack to address fact that rows that included
> the delimiter -- or a byte < delimiter -- messed up sorting in .META. Fix
> the ugly hack, and while we're at it, add a byte for flags to the HSK so can
> mark if key is a meta key, a delete key (cell/family/row), etc.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
