Dimitry Sibiryakov wrote:

>On 25 Jun 2005 at 14:20, Dmitry Yemanov wrote:
>
>
>
>>Could this work? Yeah, it requires two page writes per every backward
>>scan conflicting with a page modification, but perhaps it's
>>acceptable?
>>
>>
>
> Why can't "splitter" just acquire write lock on two pages: splitted
>and the next? Are our index reader fast enough to leave "area under
>construction" in reasonable time?
>

It means that readers block readers, destroying concurrency. It also
causes an infinite number of unnecessary writes, destroying even single
user performance.