On Mon, 11 Aug 2014 18:03:53 +0200
"Roberto E. Vargas Caballero" <k0ga_AT_shike2.com> wrote:
> Can you explain this feeling?. Maybe I am wrong, but I think Unix
> authors didn't share your opinion because listen(8) of plan9 is the
> same idea that inetd (but better done because listen doesn't need
> to parse any configuration file).
The plan 9 engineers have been the original UNIX-authors, but in many
ways, they have taken different approaches with plan 9.
The case is really simple: An event-driven tool of this kind would, if
aimed to go after the unix-philosophy, bind to specified sockets, filter
the information with certain rules and pass each on to different
sockets (like a pipeline).
Everything else (config files, service-folders, ...) is just a way to
go around the already powerful systems the os provides, so that's why
I think that inetd sucks.
---
We may discuss this even more, but the case is really simple:
Developing quark, the goal is to
1) bind to a port
2) chroot to a folder
3) serve files and cgi-requests
Providing compatibility for an event-driven network-daemon is not a
goal, as there are ways to do it with the current model by binding
quark to a "filtered" socket and let a suckless program do the
filtering.
Cheers
FRIGN
--
FRIGN <dev_AT_frign.de>
Received on Mon Aug 11 2014 - 18:32:51 CEST