![]()
| Links to HEAD: | (view) (download) (as text) (annotate) |
| Sticky Revision: |
forgot to check in some stuff after SIGCOMM deadline
keep a proper transmit event record in the flow layer so we can accurately distinguish between data and ack packets after they are sent, for better congestion control and such.
fixed a couple vegas issues
reno & vegas-style congestion control now working across multiple segments
rehashed SimLink to support asymmetric links; added semi-realistic models for DSL and cable links.
Add LGPL copyright notices, update licensing info/explanation in README
simple hacky implementation of "flow layer" starting to work
fully separate receive replay protection from acknowledgment, so that we can do late acknowledgment (e.g., for flow joining) without introducing a vulnerability.
Substantially simplified flow control based on "bytes in flight" count.
After realizing that all packets were getting queued on the transmit queue sorted in BSN order - including datagram packets having no valid BSN - I separated the transmit queues for segments and datagrams. I'm not sure I'm entirely happy with this solution though.
Major overhaul of stream IDs and attachment mechanism, to correspond with the current version of the SST paper. got stream detachment and migration working in the process (streams automatically migate to new channels as necessary, e.g., when a host's underlying IP address changes).
big reorg merge: merge -r 1807:1841 branches/reorg branch->trunk
copy sst into branch
Wrapped everything in a top-level SST namespace for cleanliness. Also, got rid of the SHA256/384/512 code now that OpenSSL has it.
Fixes and minor reorganization for Netsteria integration
Moved SST tree into proper place in SVN repository
Got prioritization working and graphed. YES!!!
Fixed bug causing packets to be permanently misplaced if the UDP transmit function returns an error...
TCP is almost working; just disconnect isn't yet...
fixed more CC-related performance bugs; seems to be working OK now.
fixed some performance bugs in congestion control, more remain...
Coded up flow support for simple 32-bit checksum armor
Got static/global data encapsulated into new Host class so it should now be possible to run multiple SST instances together.
Replaced all direct uses of QTimer with new virtualizable Timer class
Copied SST parts of Netsteria snapshot-061231, to start working on standalone Qt-SST library and test code
This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, enter a numeric revision.
| Maintained by PDOS | ViewVC Help |
| Powered by ViewVC 1.0.3 |