The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.
0.031
  - Switch to Dist::Zilla.
  - Attempt to fix tests.

0.030
  - Fix fallback store_file behaviour.

0.029
  - Shutdown writes on our socket as soon as we're done.
  - Pass through the checksum calculated in store_file.

0.028
  - Add a replacement implementation for store_file.

0.027
  - Add more documentation.
  - Handle HTTP failures better.
  - Don't repeetedly report checksum timeouts.

0.026
  - Code clean ups.
  - Handle uploading large files on 32 bit machines.
  - Handle uploading large files on 64 bit machines.
  - Handle timeouts caused by waiting for large files to be checksummed.
  - Remove the old Callback code, now disused.

0.025
  - Retag correctly.

0.024
  - Fix error path when the remote host gives you a non-200 response.

0.023
  - Don't lose create_close error messages.

0.022
  - Pass through a calculated checksum.

0.020
  - Correctly recover should create_close fail.

0.019
  - More helpful error when files grow during writing.
  - Allow a coderef to be passed rather than a dkey, for possibly safer
    re-uploading.

0.018
  - Handle things better if the call to the tracker explodes.

0.017
  - Fix bug affecting files written in multiple calls.

0.016
  - New CallbackFile client.

0.015
  - Update repository for release.

0.014
  - TCP_CORK added for the async client.
  - Don't use TCP_CORK on non-Linux OSs.
  - Try to connect to another host if the first fails (callback client.)

0.013
  - Performance monitoring for Callback writes.

0.012
  - Data can be passed by reference.

0.011
  - "Callback" subclass for when you don't have the entire file to send at
    the start.

0.008
  - Advise the OS about our file accessing through posix_fadvise(2).

0.007
  - Fix the "Connection error: unknown error" issue on first retry

0.006
  - Have a much longer timeout for HTTP response than general IO as
    lighty will move the temporary file before responding.

0.005
  - Bump timeout on read/write to 60 seconds for when machines are busy.

0.004
  - Deal with servers which reply with HTTP 1.1 to 1.0 requests.

0.003
  - Fix retrying to work correctly.

0.002
  - Add one retry to PUT request on each path given by the tracker
  - Move use of $! up to hopefully capture the socket error before it
    is removed by clearing the timeout guard.

0.001
  - Initial very rough cut