You have to provide a buffer large enough for the largest expected
packet, and you have to tell the DatagramPacket how much of the buffer
it can fill. I suppose it could use buffer.length() itself, but there
may be cases when you don't want to fill the entire buffer.

After receiving a datagram, the length is updated to tell you how many
bytes were actually copied into the buffer.

It is extremely unfortunate that the same "length" is used for both
purposes, since it is far from obvious that you need to reset the
value before reusing the same DatagramPacket to receive additional
packets (otherwise you will receive successively smaller and smaller
packets, regardless of what is actually sent).

/gordon

On Thu, 17 Nov 2005 09:49:51 +0100, Andersen
<andersen_800@hotmail.com> wrote, quoted or indirectly quoted someone
who said :

the whole idea of a buffer is to have memory ready to go to collect
the incoming characters.  If your buffer is too small, then what is
supposed to happen to the characters coming in while you are
frantically allocating more space?

In a socket you have enough buffers pre-allocated so that you can
effectively send a "pause" message to the other end before you
overflow. Even if you overflow that is not the end of the world.  In
TCP/IP can request a retransmission.

With UDP if you don't get the packet the first time, it is gone. That
is why you need a buffer ready to take all the bits without a hitch.