On Tue, Jan 01, 2013 at 09:36:29PM -0500, Corey Thomasson wrote:
> Implementing an entire userland in a library could only lead to a lot more
> sucking.
Yes, maybe it would.
On the other hand I think it can be good to use the "software tools" /
"flow based programming" approach at a smaller scale than the unix process
tree. Coroutines in C, and mini software tools similar in purpose to awk,
grep, sort, comm, join and similar, could work very well for this.
Each stream should be a stream of structs (or rather, C objects).
If it's needed to store, monitor, or send streams inter-process,
we can serialise the structs to sensible text or binary formats,
such as csv, key: value records, protocol buffers.
It might be nice if *nix processes / threads could be sufficiently
lightweight and numerous that there is no need for a smaller-scale
processing module. But I think that is not the case.
Sam
Received on Wed Jan 02 2013 - 05:22:07 CET
This archive was generated by hypermail 2.3.0
: Wed Jan 02 2013 - 05:24:04 CET