Re: CLSQL-USQL

Edi Weitz wrote:
> Why? It would work on some implementations while it wouldn't work on
> others. The same is true for parts of UFFI or various other
> cross-implementation libraries.

Well, that is not the goal of UFFI. IIRC, the only parts of UFFI which
are not supported on all UFFI-compatible implemenations are the parts
you kindly contributed which have not been yet ported to OpenMCL
because of my lack knowledge about parts of the OpenMCL FFI.

>From http://uffi.b9.com/manual/design.html#id2949258, the first
priority of UFFI is:
  * Code using UFFI must operate correctly on all supported
    implementations. 

To the extent that is not currently true in UFFI represents bugs in
UFFI.

> I personally think that the loop extensions are nice to have. Given
> SBCL's relation to CMUCL it shouldn't be too hard to make the
> extensions available there. Also, there is my code (which I myself got
> from Xanalys support) for a LispWorks loop extension in Paul Foley's
> BerkeleyDB package. This could be used as a basis for LW. So we'd have
> four implementation covered already.

Four implementations is certainly better than the 2 which are there,
one of which (SCL) is not in common usage. I think as CLSQL grows and
people want to use the most advanced features available in their
platforms, there may be more feature fragmentation due to the
differences in SQL backends. Or, like the recently added SQLite
backend, which doesn't support views or transactions. So, perhaps such
fragmentation must be accepted rather than UFFI's goal of supporting
only the "greatest common denominator" of implementation capabilities.

-- 
Kevin Rosenberg
kevin-HJRc7zDS/[email protected]