Great article, thanks!

One comment, though, which can be useful:

You mentioned the problem when you want to manually init the subscriber. So you have to stop all users, init and then get them online again.

But even in 2000, you could do it with much shorter downtime:

1.Stop users

2.Backup the database

3.Create publication and (!) subscription

4.Disable all replication agents

5.Now you can allow users to work again!

6.Whenever you can, slowly copy your backup over the WAN and restore it

7.Enable all replication agents

This way the replication will not try to copy any row until the destination is prepared. But it will store the changes in the log (you can let the log reader work so it will be stored in the distribution db instead, that way the log will be cleaned up).