

Wed Apr 10 20:15:58 MET DST 1996

	* bumped version to 0.3.10
	* added silly_rename stuff so file I/O will still work after
	  an unlink.

Thu May  2 21:09:08 MET DST 1996

	* bumped version to 0.3.11 after adding rpciod support.
	* nfsclnt is now kerneld-friendly (rpciod dies after the last
	  NFS volume is unmounted).

Fri May 10 16:21:48 MET DST 1996 (0.3.13)

	* Fixed a couple of bugs

Tue May 14 01:52:38 MET DST 1996 (0.3.14)

	* For now, I put a limit of 12 on MAX_CWND. This allows a maximum
 	  of twelve outstanding RPC requests per NFS mount, which should
	  be enough for most situations. The real fix would be to improve
	  on the cwnd computation algorithm rather than putting a clamp on
	  it. Performance is quite nice, though. With the current knfsd,
	  I get 1 MBps throughput on large files.. :-)

	* Sort requests on clnt->pending by increasing timeout. This 
	  greatly reduces the number of context switches required when
	  receiving replies. Need to implement an RTO estimator for this.

	* Now protect critical regions of code with cli()/sti(). I always
	  thought I had understood bottom halves...

