<div class="gmail_quote">On Wed, Feb 17, 2010 at 8:18 PM, Synx <<a href="mailto:plt@synx.us.to">plt@synx.us.to</a>> wrote: <blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> I tried making custom ports, but that was a mess too. I don't want or care about special values, lines, source code locations or peeking. All I want is something that consumes bytes, through a complex chain of encodings. But the chain itself has to have some way for me to "supercede" elements in it, when I no longer have to decode a prefix for instance, and all that particular saver is doing now, is just passing the bytes onto the next step forever. </blockquote><div> I am not certain if I grasp your saver and supercede concept fully, but my bzlib/port does handle the filter concept that allows you to chain a bunch of filters together on top of an input-port (and output-port too), and returns an input-port, so you can consume the data with the filters applied dynamically. For example - adding base64-encode & gzip filters on top of a regular text port, and reading from it you will get data that is first gzipped and then base64'd. The implementation does make use of thread and pipes, as well as custom ports, to take advantage of the built-in filter functions that all take an input-port & an output-port. I previously was able to run over 20K threads simultaneously on my box, so I am surprised you are having issues with just 200 threads, but I did not do additional scalability testings to see what the limit will be. If such scalability issue does exist, the implementation can be changed. HTH. Cheers, yc </div></div>