Any efficient way to select data with limit?

Question

HI
I want to select some rows of my table for showing in a webpage
number of rows is very very much and all of them cannot be showed in one page, so i used paging for my webpage
every page has just 20 rows
problem is here , how to select rows between 400 and 420 ( for example)
i am using this way of query:
select * from
select col1,col2,col3, rownum as r_n
from my table
where ... and .... and ... and ...
order by id desc
where r_n between 400 and 420
this query works, but it is slow. i think database first run inner query and select all rows then select those ones that are in my range
is there any other efficient way to do this?

Answer 1

What is a "realtime response"?
Do they really want the response to be fast on the 200,000th page? Is there actually a use case where a human being looks at 20 rows, hits next, and repeats 200,000 times? Because assuming that the user takes only a second to wait for the page to render, read the results, and hit next, that means that this one user will have spent 3333 minutes/ 55.6 hours/ 2.3 days just looking at results without a break. Are they really going to care at that point if it takes a bit longer to get the next page of results?
Part of being a developer is explaining trade-offs to the customer. If you want the database to page the results by issuing separate queries, the first page of results will be far faster than the 200,000th page. Since users are normally far more concerned about the performance of the first page than the 200,000th, this is generally a reasonable trade-off. There are alternate approaches that will likely improve the performance of fetching the 200,000th page of results. But those will generally require a more complex architecture. For example, you can return a REF CURSOR to the middle tier and let the middle tier fetch a page of results at a time. But then you have to implement the state control logic so that the right end user session sees the right cursor handle and gets the right session back (this problem is much, much easier in a client/server application)
Justin