On 2021/03/21 23:27, Florian Viehweger wrote:
> Hi ports,
>
> I've attached an updated tarball addressing hopefully all the mentioned
> issues.
>
> Thank you for your input.
In the README you have "${SYSCONFDIR}/inetd.conf", this should just be
/etc not ${SYSCONFDIR} because inetd is a system daemon not a ports one.
Also some whitespace issues in the relayd config (4 spaces then a tab,
I think maybe your editor is set to ts=4? the indent is wrong when
viewed in a normal terminal). That's in this bit:
relay "gemini" {
listen on hostname.example port 1965 tls
protocol "gemini"
forward to 127.0.0.1 port 11965
}
Please also add a "uses pledge()" marker near WANTLIB.
> > inetd, really, in 2021?!
>
> What is wrong with inetd in this use case?
Executing a new process is relatively resource intensive, most network
daemons handling short-lived requests moved away from this model in
the 80's-90's. Even the original UM gopher daemon served directly
(inetd support wasn't even present in the earlier releases, I would
speculate that maybe it was added later so people could run it under
the radar without appearing in a typical ps listing ;-)
Also explained in the old NCSA httpd docs (which eventually morphed into
Apache but this long pre-dates that change), though I suppose the bit
about parsing config isn't relevant for vger:
http://cfpm.org/docs/setup/httpd/WhyNotInetd.html
It's not a blocking thing for adding a port but still it's a bit of
a surprise since being lightweight seems to be a key point of
gemini.
