[Mono-list] System.Data

Mark Hurley debian4tux@telocity.com
Thu, 24 Jan 2002 02:43:54 -0500

On Wed, Jan 23, 2002 at 07:50:22PM -0600, Chris Podurgiel wrote:
> Everyone,
> I've spoken with Miguel about how the System.Data namespace should be implemented, and he sugested I post my thoughts on the mono-list.  According to the Mono site It was believed that we could just write a wrapper to GnomeDB.  After looking at the Namespace, I noticed that there has been a major overhaul to ADO, with new concepts like the DataSet, DataReaders, and the ability to synchronize a DataSet with an XmlDataDocument. I believe a wrapper to GnomeDB would have been great if ADO.NET wasn't an architectural overhaul from the previous versions of ADO.
> This brings up another topic of conversation.  .NET currently has two data providers: a SQL Server Data Provider, and an OLE DB Data Provider.  I'm not very familiar with Databases and Data access on the Linux/Unix platform.  Would there need to be another Data Provider for the Linux platform?

Wow...now your talking right up my alley.  I have researched the
System.Data namespace and to be more exact ADO.NET in-depth.  I can
actually *almost* tell you what MS does and doesn't do with Datasets
according to the Xml Schema standard (Xsd).  What the validating reader
does and doesn't do.  How useful (and not so useful) the new
databinding can be.

I can help code, but am very tied for at least the next eight weeks.

If anyone needs any specific information (no, I don't know it all) I
would be willing to provide additional information that I have
learned, by reading/digging and using.

I'm not familiar with GnomeDb.  But if it is like any db wrapper, it
will not solve all of the namespace requirements.

For those who are use to ADO... a good comparison is the dataset is a
container with datatables, data relations connecting the tables.  But
the dataset has no cursor functionality itself.  In ADO, the
recordset provided a storage container (sort of) and a pointer to the
current row of data.

I guess one nice side to all this is you can have multiple pointers on
datatables, dataviews, etc.

I'm not sure where the major support is needed for Linux users.
Perhaps Miguel has a choice for the database he wishes to initially
target?  Of course...the abstraction to the databases can be handled
with the Connection, DataAdapter and DataReaders.  I would aim for the
Ole style...the SQL is as basically identical, minus the performance
gains MS has coded to hit their own server directly.

Maybe create GnomeDataAdapter? What is the target, to mimick exact
names, or functionality here?

Mark Hurley