Could some point to information on the protocol used to download the host lists from host caches like router.limewire.com.
I search google but I didn't turn up much.

Thanks

It uses the normal Gnutella protocol and you dont "download" the list.
If you connect to a pongcache the cache sends you few pongs which tell you about other nodes you could connect to and the disconnects.

As easy as that

Thanks

How the pong message select the nodes I can connect to? how the selection process works? Thanks.

The selection?
I'm not sure, but I guess the pong server picks up the most efficient servers, the ones with the best connection, the most files and the biggest library (those data's are saved in the pong).

You have to ask the developers of those clients, I'm not sure since I haven't seen their code.

In its simplest form, the host cache simply returns the IP : port combination for X amount clients that connected to it previously. So, whenever YOU connect to a hostcache, your IP : port combination is stored there. The next person to connect to it, will receive your IP : port combination.

Why? Because whenever you connect to a hostcache, you pretty much indicate you need to fill one or more connection slots. The next that connects to the hostcache does too. So, what better way than hooking you two up with eachother...

But beware. The classic hostcaches are slowly being replaced by the more popular GWebCaches. GWebCaches are HTTP based scripts, which allowes anyone to install a GWebCache and thus solves some of the "centralized" aspects of hostcaches. For more information, vitist http://www.gnucleus.net/gwebcache.