dets improvements?

[Yariv Sadan]

> I don't know about losing ground. Mnesia never did target large databases,
> or even really cluster databases (even though it's turned out to be
> surprisingly competitive as a cluster database.) The standard solution for
> people handling large databases from Erlang is to use MySQL, Sybase or
> Oracle. There are lots of applications where mnesia is the obvious choice,
> mainly due to the seamless integration, but also because it runs in the
> same memory space and can serve simple requests within tens of
> microseconds.

I think that with other DBMS racing forward with clustering and soft
real-time performance, while also providing large storage capacity,
etc, Mnesia's niche shrinks. There's a good argument for using only 1
DBMS in an application, so (forgetting cost for a second) usually the
most versatile DBMS wins.

According to Claes, the dets issues that handicap Mnesia disc storage
can be fixed:
http://erlang.org/ml-archive/erlang-questions/200605/msg00042.html.
This alone will make Mnesia suitable for a much larger array of
applications. What do you think about his proposed solution?

That thread wasn't about adding ordering to dets tables. Maybe the
create_table function should accept a comparator function as a
parameter? (compare(A, B) -> first | second | equal)

>
> Having said this, I would very much like to see mnesia expand its
> capabilities, relieving more Erlang programmers from the hassles of ODBC
> and third-party product integration. (:
>

So would I :)

Best,
Yariv