What it really means is that there will be more nodes. On drupal.org the number of entries into the node table will about double. As long as nobody cares enough to provide hard test data to show Dries and us others that there is a negligible impact on performance this discussion is just hot air.
Doing 50 node loads on a single page load can be fun. In contrast, the comments get pulled out in one single query.
There are actually many more comments than nodes on drupal.org: 30172 comments, 17281 nodes. Furthermore there are 23247 issue followups which I believe are not in the comments table. Comments have a simplicity to them that makes it attractive to treat them separately. Should node access apply to comments? Should you be able to nodeapi a comment? Should comments be able to go into the book outline? Plus, you'll essentially be performing two distinct kinds of queries on that table: 1) Fetch a node, 2) Fetch a node's associated 'comments' (also nodes). Most queries would have a type = "comment" or type!="comment" check on them anyway (or "comment = true/false" if you want any node as a comment), plus a bunch of other fields and code that would be checked depending on that value. You'd be throwing together two distinct things into one big table, which is rarely a good idea. All of this sounds like a lot of work with no really good benefits in the long-run. Steven Wittens