what an fit slight paging by unequivocally infinite outcome sets sql server 2005?

edit: i'm still sharp some-more answers. thanks!

in sql 2000 days, i used temp list slight where emanate temp list new spirit buttress primary pivotal following name where spirit buttress between b.

when sql 2005 came along i found out row_number() i've controlling ever since...

but now, i found critical opening emanate row_number().
it performs unequivocally good operative not-so-gigantic outcome sets classification over an spirit column. however, it performs unequivocally poorly operative large outcome sets over 10,000 annals sorting over non-identity column. row_number() performs feeble even arrange an spirit buttress outcome set over 250,000 records. me, came indicate where throws an error, "command timeout!"

what paginate infinite outcome set sql 2005?
is temp list slight still improved case? i'm certain slight perform better... nonetheless contend there an emanate giving wrong quarrel array have multi-column primary key.

in case, i need means arrange outcome set date form column... prolongation web app.

let me know high-performing pagination sql 2005. i'd also know intelligent proceed formulating indexes. i'm suspecting selecting right primary keys and/or indexes (clustered/non-clustered) play large purpose here.

thanks advance.

p.s. does anyone know uses?

edit: cave looks something like...

select postid, posttitle, postdate
from
   (select postid, posttitle, postdate, 
         row_number() over(order postdate desc, postid desc) rownum
    mytable
   ) derivedmytable
where rownum between @startrowindex (@startrowindex + @maximumrows) - 1

postid: int, spirit (auto-increment), primary key

postdate: datetime

edit: everybody controlling row_number()?