Re: Search by urn ??

Hello Sam,

Quoting Sam Berlin <[email protected]> from ml.gnutella.dev-forum:
:See http://wiki.limewire.org/index.php?title=Mojito .  If there's
:things missing, let us know.

Yes, this page is useless to everyone but you:

	http://wiki.limewire.org/index.php?title=Mojito_Values

I want to know the low-level format, I don't care about a Java API when
implemeting a servent in a non-Java language.  What matters is the
storage representation and the architecture of the values.

Actually, I think the storage should be completely language-neutral,
and specified the same way Gnutella messages are specified, i.e. at the
byte level.

This page is incomplete:

	http://wiki.limewire.org/index.php?title=MojitoDHT_Bootstrap

How do you find the first DHT node to bootstrap and send your first FIND?

This page is only useful for internal developpers and should not be
visible from public specifications:

	http://wiki.limewire.org/index.php?title=Ping_Sequence_Diagram

Finally, this page is not precise enough:

	http://wiki.limewire.org/index.php?title=Mojito_Message_Format

For instance, it mentions things like "BigInteger".  What does that
mean in terms of number of bytes, format, etc...  Saying the BigInteger
is made of a length followed by the big integer itself, is... well...

The Contact holds a SocketAddress that can be IPV4 or IPV6.  How does
one decompile the packet and know what to expect? (again, Endianness...)

So, to conclude, it's a good start that there is some documentation,
but we are far from the quality of the documentation that was made for
other important features like the ping-pong reducing scheme, a long time
ago, or the QRP.

It is still not precise enough to start implementing anything in order
to aim at full interoperability with LimeWire.

Raphael