We should investigate whether Safari 3/Mac's implementation needs to be as complex as Safari 2.03/2.04. The Safari 2 code seems to work fine in Safari 3, so it may be best to just treat Safari the same for all of these versions. But we should still at least see how Safari 3 functions when you get rid of the history.length hack and treat it more like Firefox.

FYI, I experienced some complications with Safari 2 (involving layers of asynchronous requests) and decided to ditch 2 and support 3. I've attached a diff against svn revision 86 with the simple change I made to do this. It also works with iPhone Safari.

Patience permitting, I will investigate the Safari 2 problem and let you know if it's a bug in RSH (the app [http://eidogo.com/] works in all other browsers).

I should note that the above change still leaves Win/Safari unsupported. It may work; just didn't bother to test it.