Changed buffer-per-socket to buffer-per-operation.

This is a difficult tradeoff to measure. I think large
datagrams won't be used very frequently. So it is better to
not lock a large buffer to each socket object and instead
allocate and deallocate for each operation receiving a
datagram larger than UDP_DATAGRAMSIZE.

1 files changed, 24 insertions, 1 deletions

diff --git a/doc/socket.html b/doc/socket.html
index e6a9bf8..8a81414 100644
--- a/doc/socket.html
+++ b/doc/socket.html
@@ -90,7 +90,7 @@ of connect are defined as simple helper functions that restrict the
 
 <!-- debug ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ -->
 
-<p class=name id=debug> 
+<p class=name id=debug>
 socket.<b>_DEBUG</b>
 </p>
 
@@ -99,6 +99,19 @@ This constant is set to <tt><b>true</b></tt> if the library was compiled
 with debug support.
 </p>
 
+<!-- datagramsize +++++++++++++++++++++++++++++++++++++++++++++++++++++++ -->
+
+<p class=name id=debug>
+socket.<b>_DATAGRAMSIZE</b>
+</p>
+
+<p class=description>
+Default datagram size used by calls to
+<a href="udp.html#receive"<tt>receive</tt></a> and
+<a href="udp.html#receivefrom"><tt>receivefrom</tt></a>.
+(Unless changed in compile time, the value is 8192.)
+</p>
+
 <!-- get time +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ -->
 
 <p class=name id=gettime> 
@@ -393,6 +406,16 @@ The maximum number of sockets that the <a
 href=#select><tt>select</tt></a> function can handle. 
 </p>
 
+<!-- socketinvalid ++++++++++++++++++++++++++++++++++++++++++++++++++++++ -->
+
+<p class=name id=socketinvalid>
+socket.<b>_SOCKETINVALID</b>
+</p>
+
+<p class=description>
+The OS value for an invalid socket.
+</p>
+
 <!-- try ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ -->
 
 <p class=name id=try>