Please explain a push request

I'm trying to write a servent for fun but I am confused as to the need for the push request. My current understanding:

Gnutella servents currently listen on a high port such as 6300 or something. When you do a search and get search results, that communication is going over that port. Once you find a file and want to download it, you initiate an http transfer which happens over port 80. But the problem occurs when port 80 may be blocked for whatever reason. So the servent requesting the file sends a push request to the servent hosting the file basically saying, "hey, you have port 80 closed which means I can't grab it. Please send it to me." Then the servent with the file tries to initiate an http transfer over port 80 to send the file.

Is all this correct? Could someone please explain or point me somewhere with some good information. I've read what I have found but that hasn't been very helpful.

[Greatmerlin]

Yes, that's the way it works. The most reason a push request is sent is a firewall that denies incoming connections through port 80 on the remote host. Then this hosts initialize the transfer and sends the requested file to you. But that only works if you do NOT have a firewall that prevents incoming connections, because in this case none of the hosts can initialize the connection to the other one. So if you have a firewall, you cannot download everywhere. If you haven't a firewall you should be able to download everywhere.
But that's only theoretical. I can't say for sure it works, because I DO have a firewall.

What about port 80 though. The reason that the push request was initiated in the first place is because the "client" could not pull the file from the "server" through port 80 (standard http connection). Why would the "server" then beable to send the file OUT through port 80? Are most firewalls setup to only allow data OUT of port80 or does the server change the port that it sends the file out on?