Roman Rokytskyy wrote:

>> That's true on disk, but the in-memory internal format is fixed length
>> record buffers. So yes, arbitrary length strings would be a big deal.
>>
>
> Some time ago we were talking about improving the BLOB performance by
> fitting first 32k on the page like normal string (and handle it in memory
> accordingly) and putting the rest in "normal" BLOB structures. We could call
> this beast LONGVARCHAR/LONGVARBINARY depending on the content, and I guess
> that would fit everybody's needs. Those who are lazy enough to define the
> maximum length would declare columns as LONGVARCHAR, others would use our
> old VARCHAR columns.
>

That was my thought after some brainstorm and I would to test our
current performance of BLOB compared to VARCHAR.

This seems to be the way to go for long string support.


Adriano