dets improvements?

Yariv Sadan <>
Sat Jun 10 17:17:07 CEST 2006


>
> It certainly is, one of the real strengths of mnesia is the
> ability to have complex fields. In SQL dbs the fields are
> atomic, whereas in mnesia, they can be arbitrarily complex
> structures - trees, tuples, lists, etc.
>
> This difference alone makes all normal DB modeling theory
> ... well not mute, but everything gets easier and different.
>
> We do need to store > 2G in mnesia thoufg ... and the answer is
> not fragmented tables.
>
>

Although the need for fragmentation alone isn't a deal breaker for me
(it's more the long repair time, due to which I don't want to risk
taking my app offline for hours), it does give the impression that
Mnesia is behind the curve because no other database I know of puts
such demands on the user. Even if fragmentation happened behind the
scenes, it would be a different story. (That's what Postgres does,
actually, in 1Gb segments --
http://www.postgresql.org/docs/8.1/interactive/storage.html.)

I think Mnesia should be at least as good as MySQL and Postgres at
disc storage if not other features as well because those are the most
popular open source databases and hence Mnesia will always be compared
to them.

I'm waiting for the day Mnesia will have extensible indexing, for
instance (http://www.postgresql.org/docs/8.1/interactive/gist.html),
allowing support for full-text search
(http://www.sai.msu.su/~megera/postgres/gist/tsearch/V2/), and a
high-end query optimizer
(http://www.postgresql.org/docs/8.1/interactive/geqo.html).

Regards,
Yariv



More information about the erlang-questions mailing list