Re: [gnome-db] proposing libgda/libgnomedb as part of the GNOME platform

[Rodrigo Moya <rodrigo gnome-db org>]

On Thu, 2004-06-10 at 13:14 +0200, Murray Cumming wrote:
> I am concerned that some of the new API is not implemented by all
> providers and might never be (such as the changing of field
> definitions).
>
well, we should look at implementing them for all providers. Of course,
there would be cases where a provider is just in a bad state. In that
case, we should just make sure that provider is not compiled. But for
the most important providers (postgres, mysql, sqlite, xml, etc), we
should look at implementing all the new API. In fact, that's my goal in
the next few weeks.

>  If that will be normal then we might need some way to
> discover provider capabilities.
>
we already have, the supports() method for providers. We just need to
add new values to the GdaConnectionFeature enum.

>  Actually, I know I suggested that API in
> the first place, but I don't think it's a good idea now. I think that
> converting the data in the fields is time-intensive and lossy even if it
> a can be done. 
> 
right, the preferred way would be to run the SQL commands directly, but
that API is still a good idea, since it provides a generic mechanism to
do that for all providers.

> I am also concerned that some provider behaviour might always be
> dependent on the backend. For instance, the postgres provider can not
> report whether a field is a primary key or whether it auto-increments.
> 
it can report if a field is a primary key. Not sure about auto-
increments, but I'd say so.

> I am also starting to think that there might still be some const
> problems in the API. I will try to remove all the warnings (there are
> far too many) from libgda and maybe that will tell me more.
> 
right, we'll need a good API review before really proposing it for
inclusion.

> More importantly, there will be some resistance to putting libgda in the
> Platform before it has been used by an application in the Desktop. And
> when a Desktop application uses it, we prefer to see the library live in
> the Desktop for at least one cycle. But it's worth suggesting anyway,
> because it is unreasonable to expect a platform to also justify itself
> by providing the applications that use the platform.  
> 
my idea on proposing this is that, IMO, a good development platform
nowadays needs database access. Of course, as you say, most desktop apps
wont need a database for anything, now, but 3rd party apps using the
development platform would need a DB API. That's the rational to propose
it for the platform.

cheers