knetic / acqueduct Goto Github PK

Right now there is a lot of unnecessary overhead by assuming that each packet represents a single compressed message. Some packets are probably too small to really work this way. Instead, inflate/deflate functions (from zlib) should be used to lower this overhead.

Currently, it's expected that the user use mkfifo to create a fifo pipe for acqueduct client to read from. It would be better if acqueduct made that file itself, taking a "i" parameter which specifies the fifo location (defaulting to "./acq").

However, this would seem to make stdin no longer used to supply info to acqueduct, which is non-optimal. Best-case would be to implement an epoll system (similar to that found on the server) which takes data from either stdin or the named pipe, pushing the data through the socket no matter the source.

The server, right now, doesn't output any information about where data came from, nor does it have any format at all to differentiate packets received. It should.

Currently, zlib with the most basic settings is used to compress data. It may not be the best choice for high-throughput data. Investigation into bzip, at the very least, should occur.