Implement userland ports. (Someone could make a PhD of this.) (Prio)

[adam-7]

Implement userland ports.

Someone could make a PhD of this.

We need scalable, sandwhichable userland ports so we can implement SSL, gzip and what not as ports, with high granularity control over buffering, failure states and switching between binary and text modes.

This was brought up on the ML some years ago.

[feeley]

Abstraction and "high granularity control over everything" are contradictory. High granularity control can be obtained by using the subtypable port types that currently exist and writing all the methods they entail for the I/O operations, buffering, etc. Is this "too much control"? In other words, perhaps you would like just a little less control if it was substantially less work to implement? I guess the bottom line is: what are the I/O operations you want userland ports to support? What are the features that should come "builtin"?

[adam-7]

This is partially an open question with me. I'll intend to post the inputs I do have here in ~7-10 days' time.