I have many datafile (7.x) and some from they have from 2k to 3.5k length of
fixed part only .
For increase perfomance of row fetch,  Mr. Leonard Harvey provide me to
convert it to simple variable records.
But I don't understand - if I convert some long VarChar column to CLOB, it
restrict me to use indices with this column and take some bad issues this
ODBC/OLEDB access method's (it's documented by Pervasive).
With a right way for use both Transactional and Relation methods and  to
reach the goal ?.
For example, good performance data may take using compressed records or
split a table.

Thanks, Alex

If the data is fixed length then the record should be fixed length.
If the data is variable lenght, e.g. notes, or attachments, then the
record should be variable length with the note or attachment in the
variable length portion.

Turning on compression forces the engine to use variable length
records internally but it is different than having variable length
records.

You are correct, you can only index the portion of the record that is
in the fixed length portion.    If there needs to be an index it needs
to be in the fixed length portion.

Leonard

On Mon, 1 Sep 2003 19:00:57 +0300, "Alex D. Pichkurov"
<alex@mpc.odessa.ua> wrote:

>I have many datafile (7.x) and some from they have from 2k to 3.5k length of
>fixed part only .
>For increase perfomance of row fetch,  Mr. Leonard Harvey provide me to
>convert it to simple variable records.
>But I don't understand - if I convert some long VarChar column to CLOB, it
>restrict me to use indices with this column and take some bad issues this
>ODBC/OLEDB access method's (it's documented by Pervasive).
>With a right way for use both Transactional and Relation methods and  to
>reach the goal ?.
>For example, good performance data may take using compressed records or
>split a table.
>
>Thanks, Alex
>
>