Re: Long Type for replication

andkovacs_at_yahoo.com (Andras Kovacs) wrote in message news:<412ebb69.0309230338.2f763cb3_at_posting.google.com>...
> Okay I know that Long is type is not supported by any type of
> replication. I know that we have to use cblob instead. However some
> developpers in our office need a detailled explanation why Oracle
> doesn't support Long.
>
> Can somebody help me to explain this ?
>
> Thanks.

it was deprecated in 8.1.7.
it is still deprecated in 9.2. But there are lots of features which are not supported with data type LONG.

First thing is, it is stored inline, as opposed to out of line, as a type CLOB is. so your average row length is relavitely high, you'll see excessive row chaining, causing more blocks needing to be visited than if such data was stored out of line, diluting the density of blocks per row. If you have to use them, they might be best used as a surrogate table so that they are not stored inline.

You can use a routine whereby you store the rowid of the child LOB column in the parent row to avoid an index scan to find the child (using an insert trigger on the child table), but it won't survive an export/import.

Logical standby in 9.2 does not support LONG (and lots of other data types).
pre-oracle 7.3.4 clients cannot update it running 9.2 server. CTAS does not support it (have to use PL/SQL).

Do you want a (current) windows OS to have to support the backward compatible weak password/hashing of NTLM, leaving easy exploits, just because there are still Win95 desktops out there?

I hope your answer is no.

Ask the developers to peruse the release and de-support notes.

Oracle gave us fair warning.

Pd Received on Wed Sep 24 2003 - 00:11:38 CDT