Active development has ceased. User contributions are still welcome and considered for merging.
- changed umask to 022 (Corey Wright: umask.patch)
- relaxed timefmt constraints in http header (Corey Wright: multiple_last-modified_header_formats.patch)
- fixed long filename issue by hashing excess length
This release fixes an omission of the previous one: the inclusion of the GPL licence file. This clears the way for packaging initiatives on linux distributions and other operating systems. Other than this just minor fixes; the serious work unfortunately got lost on a crashed disk.
- added GPL licence file
- generalized fiber.py by using generator name instead of hardcoded string
- removing partial file in cache after 403 forbidden
- flushing every line in debug mode
- fix: no longer setting mtime to -1 if server does not send Last-Modified
- fix: handling empty command lines correctly
The first release in more than three years; result of my decision to give up on existing code and rewrite replicator from scratch. The new version breaks with python's built-in asyncore framework and replaces it with a much more flexible 'fiber' scheduler based on python generators. For implementation details see README.devel which is included in the package. To the user the new flexibility shows as a long new features list, which includes often requested features such as server-side download resuming, ftp support, bandwidth shaping, ipv6 support. Known issues have been solved such as the frozen download problem and the race condition that prevented joining of simultaneously started downloads. Most of the old functionality is unchanged, except for cache browsing which will be restored in a later alpha release. For basic usuage instructions please see the README.user documentation.
The server has been tested on OS X Leopard and Debian Etch, both i386 architectures. Users with access to other platforms or architectures are highly encouraged to report their findings on the http-replicator-users mailing list. Note the unit-test script which checks replicator's functionality on 16 different points, using standard gnu tools. Feature requests and general discussion are preferably sent to the mailing list as well.
- rewrite from scratch
- replaced restrictive asyncore scheduler with new 'fiber' system
- new feature: server-side download resuming
- new feature: fpt support
- new feature: bandwidth shaping
- new feature: ipv6 support
- new feature: frozen transactions killed after configurable timeout
- new feature: rudimentary off-line mode
- fixed race condition that prevented joining of simultaneously started downloads
- currently missing feature: cache browsing
- new feature: cache brower on proxy address
- new feature: client-side support for partial content
- added alias option for caching mirrors on same location
- added check to prevent access outside of cache through symlinks
- added header length restriction to fight infinite request server attacks
- created man pages for http-replicator and http-replicator_maintenance
- fixed timestamp bug; files are now properly closed before changing mtime
- suppressed size warning for chunked data
- integrated daemon code in http-replicator
- changed init.d and cron script to bash
- moved settings from configuration file to /etc/default/http-replicator
- introduced optparse module for command line parsing
- introduced logging module for output
- added support for an external proxy server
- added support for an external proxy requiring authentication
- added support for HTTP/1.1
- replicator is now suitable for maintaining a gentoo package cache
- fixed problem with absolute urls
- added posting support
- added support for servers that use LF in header instead of CRLF
- added a command line system
- fixed security issues
- improved traceback message for unhandled exceptions
- fixed problem with incomplete files after a server crash
- fixed problems with select
- fixed size calculation in cron script
- initial release.