On Mon, May 7, 2012 at 6:26 AM, Mark Wotton <mwotton@gmail.com> wrote:
probably just a spectator here, though - I haven't implemented a database in my life...

I've never implemented a database engine either,  and I suspect that describes a majority of the people here.    My wish for a relational alternative to SQL is just that:  a wish.   It seems to me that this list is likely to be far more focused on interacting with existing database engines,   like HDBC,  *-simple,  persistent, DSH, and the like.

On Mon, May 7, 2012 at 10:10 AM, Torsten Grust <torsten.grust@gmail.com> wrote:
We're thus living on the ``don't embed literal SQL text into
your Haskell source at all'' end of the spectrum, somewhat distant
from postgresql-simple and friends, I guess.  Still, we're quite
interested where these other efforts go.

Well,  I don't think embedded SQL is entirely bad,  but I don't think embedded SQL as particularly desirable either;   I see DSH as basically a higher-level layer on top of postgresql-simple and friends.   It's just that I'm mucking around in lower-level details right now, trying to get those halfway right,  so that DSH can be that much better.

Also, it seems to me that the benefits of database-specific features tend to outweigh trying to be database-independent;  the least common denominator seems terribly constraining,  and even the SQL standard is often not implemented consistently.   So one generic concern is that a higher-level abstraction doesn't get in the way of correct,  often database-dependent SQL.   But I haven't played with DSH, so I would emphasize that this is a _generic_ concern.

So I'm basically working at the problem from the opposite direction:  looking at the low-level interfaces and the pain points that they cause and trying to improve those.   Another obvious problem is failing faster:  it'd be really nice to have syntax checking and end-to-end typechecking;   although this can't always be done 100% statically because there are reasons to use dynamically-generated SQL. 

Best,
Leon