I've recently made similar descisions, and ended up building an HTML table
because all my rows we're not the same.  I've also written paging code for it
as well.  If all of your rows are the same, or you want to add additional
logic, then you may be better off with a repeater.  If you do go down the
HTML route, I have a few pointers which will help.

If you want an event to fire from your table (e.g. you've got a link button
in a cell), then you need to create the table and the link buttons on every
load.  You can't just build the rows and not bother on a post back if you're
not displaying them.  To do this, I recommend that you have two functions - a
PrototypeTable() function and a PopulateTable() function.

In PrototypeTable to create the TableRows, TableCells and any Labels or
LinkButtons or other objects you'll need.  You can set any event handlers
here as well.  Set all the TableRows to invisible.

In PopulateTable, set the values of the labels and other objects, and set
the visibility.

The advantage of doing this is that you know what objects you have in your
program.  You know you've always got ten TableRows called 'salesTableRow1' to
'salesTableRow10', and you can control the events.  When you press the paging
controls, you get a postback which runs PrototypeTable(), sets the start
index for your data and then runs PopulateTable() which will set the correct
values in the labels.

As I said earlier, a repeater is almost certainly easier to work with, but
if you want to create tables then this could save you a lot of time