FreeTDS Development Group

[Michael Peppler <mpeppler AT peppler.org>]

On Tue, 2002-05-21 at 14:22, Lowden, James K wrote:
> > From: Michael Peppler [mailto:mpeppler AT peppler.org]
> > Sent: May 21, 2002 4:59 PM
> > 
> > On Tue, 2002-05-21 at 13:45, Lowden, James K wrote:
> > > The assertion is that MS SQL Server 2000 has no support
> > > for 4.2 (or, of course 5.0) connections.  Is that in fact true?  Can
> > > Sybase's Solaris ct-lib client connect to MS SQL Server 
> > > 2000, or not?  
> > 
> > AFAIK this is correct. However, DBlib apps may still work.
> 
> Why would db-lib apps work?  I thought it was a TDS issue.  Possibly having
> to do with dropping ascii support at the protocol level (in favor of UCS-2).

It may have something to do with the initial login packet. The
non-patched MS-SQL 7.0 server wouldn't accept a CTlib connection, but
handled DBlib connections OK.

> > > FreeTDS is a possibility, but its bcp logic would have to 
> > be amplified to
> > > include host variable and transaction support.    If you 
> > think it's really
> > > the best answer, though, that's information, too.  
> > 
> > Well - bcp is not transactional (or at least not really), and doesn't
> > support host variables (at least not if I understand what you mean by
> > that).
> 
> I see.  You're saying bcp is "not transactional" because the data don't hit
> the tranlog, so there's no rollback/rollforward and such for recovery and
> backup.  I should have said "bcp has to support batches".  

Ah - yes, I see.

> Right now, FreeTDS writes the whole file as one batch.  If there's a glitch
> in the data, the batch is not committed, and in fact "sp_who" will show the
> spid's status as "rollback".  In such a case, process will remained blocked
> for an inordinately long time, much longer than it took to upload the rows.
> So, we need bcp batch support.  :)

Absolutely.

> By "host variables", I mean feeding the bcp API with bound memory variables,
> rather than from a file.  You know, bcp_bind, bcp_sendrow, and friends.  

Yep - I was definitely on a different page :-)

Yes, to make a bcp API really useful these would be necessary.

> Thanks for the reply, Michael.  Your name came up, you won't be surprised to
> hear.  

:-)

Michael
-- 
Michael Peppler                              Data Migrations, Inc.
mpeppler AT peppler.org           *or*          mpeppler AT mbay.net
http://www.mbay.net/~mpeppler
International Sybase User Group: http://www.isug.com

Attachment: signature.asc
Description: This is a digitally signed message part