On 01/04/07, Frames Project <frames(a)lct.jussieu.fr> wrote:
1. why the innoDB type is chosen for the tables except
InnoDB is better; it supports larger tables, longer indexes, and has
decent support for transactions, which is somewhat essential to keep
data consistent.
hitcounter (HEAP)
This table is used to temporarily stash the results of view counters
per-page. On a periodic basis, this information is "flushed" back into
page.page_counter. Using an in-memory table for this is faster. The
information here isn't so critical.
searchindex (MyISAM)
The default search engine uses full text matching, which is a feature
of MyISAM, hence, the search index table has to be MyISAM.
Is there a way to ask for MyISAM type for all ?
It's not advisable.
2. why I got latin1_swedish_ci for *collation* and
not some utf8
character_set instead, did I missed some parametrization?
You selected the defaults. MediaWiki's current default behaviour is to
ignore the collation of the database, and store UTF-8 data in latin1
tables; it doesn't care what MySQL thinks the collation is. This is
because MediaWiki predates MySQL's shaky UTF-8 support (which still
isn't quite up to what we need).
This can cause problems if MySQL tries to perform conversions on
import/export; see the mailing list archives for information on
getting around this.
We provide two experimental schemas which can be selected during
installation of 1.9.0+ that use UTF-8 collation, and configure
MediaWiki to treat the database server as using UTF-8. These are
experimental.
Rob Church