cvs file checksum ?

Matthew Dillon dillon at apollo.backplane.com
Tue Apr 4 09:46:45 PDT 2006


    Maybe we should just write our own client/server for cvs tree
    synchronization.

    It actually wouldn't be too hard.  We don't have to do anything fancy
    to be able to support incremental updates for 'most' people.  We just
    need to keep around diffs of the actual cvs file(s) for each commit.
    The client supplies a MD5, the server matches it up against either the
    file in the CVS repository or a backpatched version of that file and
    then just transfers the patch(es).  If the server can't match it up,
    the whole file is transfered.

    Nothing fancy at all.  The server wouldn't even have to be written to
    understand or parse CVS.

						-Matt






More information about the Kernel mailing list