[freetds] Ethereal/TDS development

ZIGLIO, Frediano, VF-IT Frediano.Ziglio at vodafone.com
Tue Jan 2 11:38:10 EST 2007


> 
> On Fri, 22 Dec 2006 22:49:56 -0500, "James K. Lowden" (in 
> FreeTDS Digest, Vol 47, Issue 27) wrote:
> 
> > Brian Bruns did the original and afaik only work to teach 
> TDS to Ethereal.
> > I wouldn't expect him to contribute further to it.  Someone 
> else will
> have to pick up the ball.  
> 
> James, thanks for the datum.
> 
> Hmm... not saying that I am equal to the task, but if I were 
> to try to "enhance" TDS/Ethereal, there are a couple of basic 
> TDS 5.0 questions I have:
> One of the things that has bugged me about Ethereal's 
> interpretation of TDS 5 is that it sort of stumbles when it 
> encounters a TDS_KEY (client side) or TDS_ROW (server side).  
> To resolve this, I assume I would have to hunt backwards 
> through the stream, for the related ROWFMT(2) and/or 
> PARAMFMT(2).  Sybase's description of the cursor tokens makes 
> this challenging: apparently you pass the PARAMFMT after the 
> CURDECLARE, and the PARAMS in the CUROPEN, to which the 
> server responds with a ROWFMT accompanying the CURINFO 
> (describing the subsequent ROWs and KEYs). All of these 
> associations appear to be inferred, as there are no 
> identifiers relating one to the other.
> Can I assume that only one ROWFMT is "active" at a time?

No, every cursor can have a different format and there are normal rows
too.

> Can I assume that if the PARAMFMT does not immediately 
> precede a PARAMS, I should use the PARAMFMT supplied with the 
> most recent CURDECLARE?
> 

I think not, you can have multiple cursors active. This is the good
thing with cursors (but I have to admit the worst for tracing)

> Thanks for any information.
> 
> (still no news on "SQL Anywhere CMDSEQ"?)
> 
> C.
> 

bye
  freddy77



More information about the FreeTDS mailing list