- make partitioning support multiple engine types in one table (compressed MyISAM and InnoDB)
- make row-based replication binlog events optionally include the original SQL statement
- make the
--base64-output=DECODE-ROWS option for mysqlbinlog include column names
I have done some QA and performance testing with it and it was solid. I look forward to using the InnoDB plugin with it. Maria and Falcon would get more early users if they were to release plugins for 5.1. That would also validate the handler plugin interface, as in it is pluggable when a new storage engine does not require a new MySQL release.
Posts
This really did not sounds like a strong endoresment.. http://monty-says.blogspot.com/2008/11/oops-we-did-it-again-mysql-51-released.html
ReplyDeleteIt was more like "It sucks less than 5.0, if you don't use any new 5.1 only features."
ReplyDeleteUnfortunately, partition and ndbcluster violate the pluggable storage engine model by digging into mysql internal data structures - so it would be non-trivial to have a third-party partition engine. Worse yet, several data structures change if the partition engine is compiled in or not. IMO, those ifdefs needs to go away so that the ABI of MySQL 5.1+ becomes 'stable' enough for plugins to work on at least the next release.
ReplyDeleteIf community would solve the issue with
ReplyDeletepartitioning supporting multiple table
types I would be thrilled. The tricky
part with this is the fact that the
storage engine API can only show up
one face of a table, so e.g. it is
either transactional or not, and
combining MyISAM and InnoDB means it is
both. Same with parameters like little_endian
and compressed or not (which at least a
few years ago was present in the table
share info)
To Antony's comment:
Hi by the way Antony, hope your well
partitioning handler isn't really intended
to be a pluggable engine. However it's
easy to not use the partition engine if
your third party engine wants to be a
partitioning engine, there is a handler flag
for this and if the handler sets that it
assumes that the storage engine wants to
do partitioning natively as NDB does.
I *did* endorse 5.1 as a better 5.0 if you only use 5.0 features. 5.1 features are starting to be ok (they work for a lot of people), but I would not say that they are of expected GA quality as there are still known fatal bugs in most of them.
ReplyDeleteMark Callaghan wrote, “Monty endorses MySQL 5.1″, a claim that some of his readers dispute. -- Log Buffer #126
ReplyDeleteI took a look at his post about the bugs that come along with it. Maybe add those into your post. Thanks for the follow up though.
ReplyDelete