Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-01 00:23:12 -0700 (Fri, 01 Jul 2005)
Revision: 1850
Log message:

      *** empty log message ***
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 00:57:00 -0700 (Tue, 05 Jul 2005)
Revision: 1851
Log message:

      Added some files from the previous implementation of the group communication
      with token passing. The test program doesn't compile at this point (flooder)
      but I'm working on it. I wanted to have the rest of my changes in because my
      home HDD started making some weird noises.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 15:52:15 -0700 (Tue, 05 Jul 2005)
Revision: 1852
Log message:

      This code compiles but it does nothing, so far.
      I need to write the send function next and see what happens.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 15:53:14 -0700 (Tue, 05 Jul 2005)
Revision: 1853
Log message:

      Ooops... Forgot to add this file. Now it should compile fine.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-05 16:34:32 -0700 (Tue, 05 Jul 2005)
Revision: 1854
Log message:

      *** empty log message ***
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-05 16:36:47 -0700 (Tue, 05 Jul 2005)
Revision: 1855
Log message:

      Nothing useful yet. Just creating the directory structure and signatures.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 16:46:45 -0700 (Tue, 05 Jul 2005)
Revision: 1856
Log message:

      Some structural changes. Very little code addded.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 16:48:57 -0700 (Tue, 05 Jul 2005)
Revision: 1857
Log message:

      Needed to commit these too for the whole project to compile.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 17:01:20 -0700 (Tue, 05 Jul 2005)
Revision: 1858
Log message:

      Renamed the files again.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 17:03:20 -0700 (Tue, 05 Jul 2005)
Revision: 1859
Log message:

      Forgot to add another file...
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-05 18:46:09 -0700 (Tue, 05 Jul 2005)
Revision: 1860
Log message:

      More changes. It still isn't functional but it's getting warm.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-06 13:57:05 -0700 (Wed, 06 Jul 2005)
Revision: 1861
Log message:

      first try at dio.mli
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-06 13:57:22 -0700 (Wed, 06 Jul 2005)
Revision: 1862
Log message:

      add omakefile
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-06 13:58:03 -0700 (Wed, 06 Jul 2005)
Revision: 1863
Log message:

      add dio to omakefile
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-08 14:49:42 -0700 (Fri, 08 Jul 2005)
Revision: 1864
Log message:

      Added naming parameters in naming_params
      Start defining the naming-service client in lnsc
      A few other minor changes.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-08 16:07:13 -0700 (Fri, 08 Jul 2005)
Revision: 1865
Log message:

      beginnings of DIO, though functors still confuse both me and ocaml
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-08 18:42:47 -0700 (Fri, 08 Jul 2005)
Revision: 1866
Log message:

      Changed Id.t from abstract to string. DIO uses Id.t for its chunkids
      (actually group IDs since IIO does the translation) and IIO has
      to be able to generate an Id.t from its own chunk IDs. If Id.t were abstract
      the only option would be to add an ID.dio_id_from_iio_id function that would know
      how to translate from IIO chunk ID's to DIO ID's.. seemed like something far
      too specific to put in the ID module.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-08 19:14:13 -0700 (Fri, 08 Jul 2005)
Revision: 1867
Log message:

      Placed IIO types that LNSC needs to know about in a separate file: iio_private_types;
      couldn't find another way of sharing types between IIO functor and LNSC module.
      Added more function definitions and comments to iio modules.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-10 18:53:14 -0700 (Sun, 10 Jul 2005)
Revision: 1868
Log message:

      Commited some of the things I've been working on over the past few days:
      - the new gc timer module
      - the implementation of the sendmsg function in C/ML
      - rewrote some of the gc code, still not completly implemented...
      - added a couple of files Vlad forgot to add in his commit.
      - removed dio from the .SUBDIRS in main OMakefile since the code in the
         dio directory does not compile.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-11 12:43:41 -0700 (Mon, 11 Jul 2005)
Revision: 1869
Log message:

      Moved group_locator type definition from gc/type_defs.ml to common_types.ml
      since IIO, DIO and GC all need it. All GC code compiles, so I think there is
      no problem
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-11 15:39:13 -0700 (Mon, 11 Jul 2005)
Revision: 1870
Log message:

      After talking to Cristian: Id.t remains abstract, for now both IIO and DIO
      use the same type for their types; if translation required we add this later
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-12 15:00:27 -0700 (Tue, 12 Jul 2005)
Revision: 1871
Log message:

      Getting closer to actually compiling, the main problem seems to be the type
      of "data" between DIO and GC. Not sure how to make this actually work in the
      proper ocaml fashion.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-12 16:48:03 -0700 (Tue, 12 Jul 2005)
Revision: 1872
Log message:

      Added setbyte function (required by IIO to generate metachunk ID from
      datachunk ID prefix)
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-12 17:00:57 -0700 (Tue, 12 Jul 2005)
Revision: 1873
Log message:

      After previous commit interface was not uniform: in byte function, a byte
      was represented as type char whil in setbyte it was an int;
      changed setbyte to receive byte as char as well
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-12 17:45:57 -0700 (Tue, 12 Jul 2005)
Revision: 1874
Log message:

      First version of datachunk lookup  (in module Lnsc).
      Comments:
      -> right now only splitwhenfull metachunk split policy considered;
         at first sight seems easy to also provide splitoncollision support here
      -> most part of the code is commented out; it can't compile without a Dio
         module
      -> even if it compiled, at this stage it would be hard to test it; for that
         we need to get to the point where we have a hierarchy of metachunks and
         datachunks (maybe a simulated one...)
      
      are commented because it can't compile as long as we don't have Dio
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-13 02:59:14 -0700 (Wed, 13 Jul 2005)
Revision: 1875
Log message:

      A few more changes to gc.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-13 15:42:07 -0700 (Wed, 13 Jul 2005)
Revision: 1876
Log message:

      A few makefile changes.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-13 15:58:04 -0700 (Wed, 13 Jul 2005)
Revision: 1877
Log message:

      DIO compiles cleanly now
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-13 15:59:54 -0700 (Wed, 13 Jul 2005)
Revision: 1878
Log message:

      I don't think we need to set the data type to be iovec in GC, isn't that why we have a functor in the first place?
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-13 16:38:09 -0700 (Wed, 13 Jul 2005)
Revision: 1879
Log message:

      DIO's data type is now an iovec
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-13 16:48:14 -0700 (Wed, 13 Jul 2005)
Revision: 1880
Log message:

      Added dio to SUBDIRS in main makefile.
      Made type data iovec in gc.mli
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-13 18:47:18 -0700 (Wed, 13 Jul 2005)
Revision: 1881
Log message:

      More changes. Included one function from ocaml (get_sockaddr) in the clib.
      Created a new flag to be used with sendmsg (MSG_DONTWAIT). More minor changes.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-14 15:26:08 -0700 (Thu, 14 Jul 2005)
Revision: 1882
Log message:

      Created a sendQ for the main thread and added the appropriate helper functions
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-14 15:47:24 -0700 (Thu, 14 Jul 2005)
Revision: 1883
Log message:

      Renamed gc to gcomm and fixed the files using it.
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-14 16:07:33 -0700 (Thu, 14 Jul 2005)
Revision: 1884
Log message:

      -> byte is now represented as int everywhere
      -> lnsc is no longer used, code used to locate chunks is now in iio.ml
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-14 17:56:43 -0700 (Thu, 14 Jul 2005)
Revision: 1885
Log message:

      *** empty log message ***
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-14 18:27:23 -0700 (Thu, 14 Jul 2005)
Revision: 1886
Log message:

      Implemented IIO Read and Write.
      Code is still not tested.
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-14 18:44:55 -0700 (Thu, 14 Jul 2005)
Revision: 1887
Log message:

      Wrote the send function. However, there is a bug in the c library and the
      code segfaults... need to check that next.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-15 15:36:11 -0700 (Fri, 15 Jul 2005)
Revision: 1888
Log message:

      add some helpful comments
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-15 16:40:51 -0700 (Fri, 15 Jul 2005)
Revision: 1889
Log message:

      Fixed the segfault. Now on to the rest of the code...
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-15 17:55:29 -0700 (Fri, 15 Jul 2005)
Revision: 1890
Log message:

      Added capability of storing chunks to disk to DIO.
      Also, gave DIO its own dsc type, and reworked the internals
      a bit. I knew it looked too easy before.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-15 18:12:21 -0700 (Fri, 15 Jul 2005)
Revision: 1891
Log message:

      add delete_chunk
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-15 23:13:25 -0700 (Fri, 15 Jul 2005)
Revision: 1892
Log message:

      adjust some types in dio to be more correct
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-16 14:44:03 -0700 (Sat, 16 Jul 2005)
Revision: 1893
Log message:

      *** empty log message ***
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-16 14:52:44 -0700 (Sat, 16 Jul 2005)
Revision: 1894
Log message:

      Added locatefmchunk - will be used by create to locate the final metachunk where locator for a newly generated data chunkid should be stored; for now only splitwhenfull policy considered
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-18 14:15:21 -0700 (Mon, 18 Jul 2005)
Revision: 1895
Log message:

      
      More fixes. For now I use "brute force" marshalling/unmarshalling but this
      should change in the final version. Still need a couple of fixes in the
      functions processing the internal messages.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-18 15:04:45 -0700 (Mon, 18 Jul 2005)
Revision: 1896
Log message:

      fix dio OMakefile
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-18 15:16:58 -0700 (Mon, 18 Jul 2005)
Revision: 1897
Log message:

      change GC to take a proc_info local address instead of a port to bind to
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-18 15:34:26 -0700 (Mon, 18 Jul 2005)
Revision: 1898
Log message:

      DIO can now see when it gets its own message back
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-18 16:08:26 -0700 (Mon, 18 Jul 2005)
Revision: 1899
Log message:

      add set_global_timer, with dummy implementation, are the types right?
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-18 17:01:07 -0700 (Mon, 18 Jul 2005)
Revision: 1900
Log message:

      dio now saves modified chunks with a global timer
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-19 01:15:58 -0700 (Tue, 19 Jul 2005)
Revision: 1901
Log message:

      More significant additions:
      -acks
      -retransmitQ
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-19 04:08:20 -0700 (Tue, 19 Jul 2005)
Revision: 1902
Log message:

      More changes. Now the flooder takes the right parameters.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-19 14:29:25 -0700 (Tue, 19 Jul 2005)
Revision: 1903
Log message:

      go back to saving each chunk to disk on each write, if we're a replica
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-19 16:14:48 -0700 (Tue, 19 Jul 2005)
Revision: 1904
Log message:

      add a rudimentary test program for DIO
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-20 15:46:43 -0700 (Wed, 20 Jul 2005)
Revision: 1905
Log message:

      add tests of chunk store to DIO's test suite
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-20 16:06:22 -0700 (Wed, 20 Jul 2005)
Revision: 1906
Log message:

      Added chunk locking related functions and type definitions.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-20 17:04:48 -0700 (Wed, 20 Jul 2005)
Revision: 1907
Log message:

      * move DIO's configuration parameters into dio_params.ml, as per
        Vlad's suggestion
      * fix small bug in save_chunk
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-21 16:37:40 -0700 (Thu, 21 Jul 2005)
Revision: 1908
Log message:

      fixed small bug in chunk store and added a test
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-21 17:13:48 -0700 (Thu, 21 Jul 2005)
Revision: 1909
Log message:

      More code added to support the token passing protocol. The code currently
      compiles but I didn't have a chance to test it.
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-22 16:11:47 -0700 (Fri, 22 Jul 2005)
Revision: 1910
Log message:

      fix DIO based on my newly corrected understanding of what an iovec is
      

Changes by: vgrama (vgrama at unknown.email)
Date: 2005-07-25 10:18:41 -0700 (Mon, 25 Jul 2005)
Revision: 1911
Log message:

      Skeleton for create implementation. Doesn't handle:
       -> working with the root metachunk
       -> chunk split
       -> selecting replica holders for the new chunks
      

Changes by: arcady (arcady at unknown.email)
Date: 2005-07-25 13:47:07 -0700 (Mon, 25 Jul 2005)
Revision: 1912
Log message:

      fix compile breakage
      

Changes by: Cristian Tapus (crt at cs.caltech.edu)
Date: 2005-07-31 01:55:30 -0700 (Sun, 31 Jul 2005)
Revision: 1913
Log message:

      More code that compiles. This should be most of the code needed by GC.
      It is missing a smart marshaling of messages. Currently it's implemented
      using OCaml's Marshal module.
      
      The flooder program can be used to test the code. I only tested a very
      simple setup with just two nodes and one group. More tests are necessary
      to guarantee the code is fully functional...
      
      This version of the code has a LOT of debug output messages. The output
      will most likely not make sense to anybody else but me at this point, but
      I'm too tired at this point to clean that up. I'll try to do it in the next
      couple of days.