On 08/29/08 00:29, Dag H. Wanvik wrote:
Thanks for the overview of this issue, Kristian. I agree this looks
like something we need to address; LOBs are much used and good
performance here is important.
Thanks for the feedback, Dag.
I won't have time to work on this right now, but hope to do some more
work later. People should post their comments if they have any.
Kristian Waagan <[EMAIL PROTECTED]> writes:
I'm bringing up b) because JDBC has methods for inserting data without
specifying the length. Throwing out some ideas:
- obtain required information on the fly, update immediately after
insert is done
This seems the obvious choice; what is the downside, if any, to this
option?
Nothing in particular that I know about yet. As long as we make the
number of bytes (possibly) used for length information fixed, it should
be fine to update them after all the data has been inserted.
--
Kristian
Storing both lengths seems cheap enough for large LOBs; if we worry this is
an issue for smaller LOBs, I guess we could encode the ints, so short
lengths take little space..
- insert data without length information, then
* update on first subsequent request
* use a background task to update information
* add a "maintenance routine" for updating the information
