changes.svn-base

网络模拟器

Mon Mar 22 19:06:29 GMT 1999 - (Chris)
  - Merged Alex's code in
  - Added IP to bogus (NAM and Javis) node number translation. Linear search
    but who cares
  - It works. I just witnessed the lovely three-way handshake of the go back
    n protocol...
  - Added a new example, Test_Bigger .. Creates a star-type topology with a
    router in the middle and uses SGN to send packets around.
  - fixed the Test_Packet example a bit..
  - Added yet another test/example, Test_LinkDown.java
  - Updated Events.sgml and Structure.sgml.
  - Added data transfer to SGN. Looks nice with window size 4 and it
    actually sticks to it. However, out of sequence delivery would break it
    (there's no timeout or anything)
  - 

Fri Mar 19 10:42:51 GMT 1999 - (Chris)
  - Added more tracing support. Nodes attach their IP handler to a trace
    object now, it used to be hidden from tracing..
  - Added attachWithTrace to the Simulator class that will attach elements
    and do tracing setup at the same time. Saves some code.

Wed Mar 17 10:08:29 GMT 1999 - (Chris)
  - Fixed the agent scheme a bit so it uses a more realistic approach: The
    new 'recipe' for sending packets is:
      1) Call the 'canSend()' function of the lower-level agent
      2) If yes, call the 'send()' function of the lower-level agent
      3) If no, stop trying. Wait for a call to your indicate function with
         the 'READY_TO_SEND' flag set. Start again at 1).
  - I have decided that the SGN protocol will always accept packets and
    buffer them.. so canSend returns true, always.
  - Links do not accept packets when they're down anymore
  - Added a 'LinkStateCommand' that can be used to bring up or take down 
    links at specific times.

Tue Mar 16 22:34:18 GMT 1999 - (Chris)
  - The experiment continues, it's 22:34:18 (see below).
  - 00:26:56. Blatantly ripped the TCP finite state machine for the SGN
    protocol. All the connection setup and teardown is in there now.
  - I have no means of testing all this so I've decided to write two
    classes, RandomSource and RandomSink that will set up a SGN connection
    between each other an send a random amount of data at random time 
    intervals.
  - Fixed a big in the IP handler. The protocol field did not get set ->
    couldn't send any SGN packets because they didn't get passed up.
  - 01:38:43. Connection established. Ok, I have now spent .. 4 hours on 
    this protocol and I have never written a connection-oriented protocol
    before let alone read about the details. Not too bad. I'll do data
    transfer tomorrow. (we're over 4000 lines of code now, more to come)
  - Moved the Route class outside the RoutingTable class, people should be
    able to see it.
  - Added two members to the route class that could be potentially useful  
    for people writing routing protocols:
        Object metric - a metric associated with this route, e.g. hop count
        int ttl - time to live. This could be decremented and the route 
                  deleted when it reaches zero
    Both are not used by the IP implementation but really for people who 
    want to extend the routing
  -

Mon Mar 15 21:54:34 GMT 1999 - (Chris)
  - Links have a status up/down now. If set to down they drop all packets
    on the immediately
  - TIMING STARTS. I am implementing the go back n protocol now and timing
    how long it takes. It is now 22:11:22.
  - HALF FINISHED. It is 22:33:25 and I have finished the port multiplexing
    class (SimpleGoBackN) that deals with handing packets to the right
    ports and the SimpleGoBackNAgent object sitting there. 
  - 22:45:00 can't continue, have to do other coursework..

Sun Mar 14 21:57:44 GMT 1999 - (Chris)
  - Added higher-level protocol handling to IP Handler.

Thu Mar 11 01:01:05 GMT 1999 - (Chris)
  - No coding today, the report is due soon
  - But I wrote Overview.sgml, the file that gives insight into the internals
    of JNS and hints on how to extend it..
  - Also finished Structure.sgml with example source code. The manuals are now
    >5000 words in total, not too bad.. but they're not complete still

Tue Mar  9 21:43:13 GMT 1999 - (Chris)
  - Finished IP fragment reassembly, looks fine. It works with the test 
    example (which does not deliver out of sequence.. but the code is
    written to work out of sequence, it's just not tested yet..)
  - Fixed up the whole business of link-to-interface packet transfer a lot,
    the link indicates when it is free now
  - Added the first bits of tracing. Events get generated when packets are 
    sent and when packets are enq'd now (Of course, noone's writing them yet)
  - Wrote jns.util.EventGenerator, a set of utility functions that make
    event generation easier
  - More tracing, queues send enqueue, dequeue and drop events
  - Wrote a documentation file that strictly prescribes the essential
    parameters of Packet Events (e.g. HopEvent, ReceiveEvent, etc.)
  - Added javadoc comments to a number of source files

Mon Mar  9 23:59:04 GMT 1999 - (Chris)
  - The first IP packet just made a whole round-trip around a network, time
    to celebrate.. Well, the network consists of just three nodes (two hosts
    and one router) but it shows that routing seems to work.
  - I had to remove automatic creation of default routes when attaching
    interfaces because it creates a loop when used with routers who cannot
    route a packet.
  - Implemented IP fragmentation, sort of works, but seems to generate a lot
    of events. Certainly needs to be checked. Reassembly is missing..
  - The many events are generated because the link sets a flag to 'cannot 
    send' as long as one packet is being put on the wire, so it's fine. The
    interfaces just retry very frequently.

Sun Mar  7 21:33:03 GMT 1999 - (Chris)
  - Added processing of incoming packets to interface and IP Handler, so
    there is support for nearly a complete round trip now. 
  - I spent the day writing about this instead of coding on it, so there's 
    hopefully going to be more tomorrow..

Sat Mar  6 12:52:49 GMT 1999 - (Chris)
  - Wrote a PriorityQueue class to store the commands to be processed by the
    simulator, where the priority is the time of the command..
    Also wrote a test program.. TestPriority.java
  - Implemented SimplexInterface to SimplexLink and DuplexInterface to 
    DuplexLink attach functions. Harder than it sounds! there were lots of
    checks to do.. 
  - Two new test programs: TestSimplexAttach, TestDuplexAttach. Nuff said ;)
  - Added the 'Route' and 'RoutingTable' class, everything in there to do
    some routing, apart from the deleteRoute functions are not implemented.
    (Don't think anyone would use them anyway)
  - Added TestRouting.java 
  - Lots of small additions to IPHandler, QueueDropTail
  - IPHandler routes packets now. There is a flag whether the first
    interface added to a node will become a default route for packets.
  - Moved the whole lot to a jns/ directory, necessary to properly import
    all this when it's in a jar.
  - Implemented a few update functions. Indeed, packets get sent to the next
    hop now, they go on the interface, queue, link, but they don't get taken
    off yet. Routing works, too.
  - Wrote Structure.sgml, describes how to build the static network 
    structure from JNS components
  - Updated the build_docs script to allow building of latex or HTML
    documentation as an option

Thu Mar  4 22:28:02 GMT 1999 - (Chris)
  - Back from the Comms&Nets blues
  - Added SimplexLink/DuplexLink
  - Implemented the DropTail queue.. about 10 lines of code in total, I'd
    say, without update()
  - Wrote the util.Queue class because Java doesn't provide one. Use for 
    anything that might require a queue (e.g. IPHandler for outgoing
    packets)
  - Added the command/ classes. You can specify now what happens when in 
    the simulator. Also added 'StopCommand', which will stop the simulator..

Wed Mar  3 01:02:34 GMT 1999 - (Chris)
  - Nope, didn't miss a day, it's just past midnight :)
  - Brushed up the elements/ classes a bit, added functionality, filled in 
    bits
  - Added some more Javis tracefile preamble writing. Biggest TODO at the 
    moment: IP Addresses have to be translated to simple numbers for Javis/
    NAM.
  - Wrote the 'ExtendJavis' SGML file in doc/.

Mon Mar  1 23:46:00 GMT 1999 - (Chris) 
  - Rough night tonight, wrote about 1500 lines of code outlining the
    simulator.
  - I added all of the classes in all the subsystems.
  - Wrote two little test programs, one testing the structure and one will
    one day test sending a packet.
  - The Javis trace file writer already writes a preamble :)
  - Added the usual Makefile I always add, with the following functions at
    your disposal:
	'make clean' - get rid of all .class files
	'make wc' - Do a word (and line) count on the .java files
	'make rmtilde' - get rid of all backup files (*~)
  - Added a Preferences class