At 09:55 PM 12/7/2001 -0400, Claudio Valderrama C. wrote:

>No, there are C #ifdef preprocessor instructions surrounding code that's
>mean to be compatible with GATEWAY. Gateway has nothing do to with the PC
>seller but with an old "gateway" from IB towards other (now forgotten)
>products like RDB/ELN and other RDB??? flavor (database engines, if you
>don't follow).

Not quite. The rdb*.c modules handle communication with Rdb/VMS and
it's family. GATEWAY connects to Oracle itself. Or at least a version
of it. The code could (one supposes) be resurrected and extended to
MsSQL etc.

>The code that pads strings with blanks over the wire is another example,
>since we had DecNet, Banyan Vines, TCP, IBM SNA... and the engine had to
>support several of them (not SNA to my knowledge). Now, TCP rules.

Not entirely correct either. Records are passed around internally in
fixed length buffers - thus with varchars expanded. Our assumption was
that the remote transfer protocol would compress data if compression
was desirable. That still seems reasonable to me - the transport code
knows a great deal more about the relative cost of sending large packets
vs compress/decompress. However, it's not done, so I guess we should
look at doing it. My recommendation is not that we truncate varchars
but that we use the existing sqz (squeeze) code and compress everything.

And that we measure the result before releasing it.


Regards,

Ann
www.ibphoenix.com
We have answers.