At 08:49 AM 10/4/2004, Arno Brinkman wrote:

>IIRC blobs aren't compressed.
>Beside your solution it would be usefull to compress blobs (that fit into
>data-page) also.

The ordinary runlength compression doesn't work well
on blobs because they don't, typically, have long runs
of identical characters. Other compression algorithms
would work, but introduce a different level of cost
and complexity. Blob filters can easily provide
compression and can be tuned to match the type of data
expected.

Jim's message suggested (inadvertently) that there was
a flag in the header that caused the system to store blobs
on separate pages. What he meant is that there is a bit
on a page header that says it contains a large object -
blob or record fragment - and should not be scanned during
a natural table walk.

Regards,


Ann