On Mon, Mar 18, 2013 at 7:52 AM, Tomasz Torcz <[email protected]> wrote:
> Hello,
>
> I'm a Fedora packager and OWFS user. I thought about getting OWFS into
> Fedora proper. I'll base my work on .spec file shipped with OWFS, but
> I'd like to make it Fedora specific.
>
> Great news.
> As you may know, Fedora is one of few distributions using systemd as
> init and service manager system. Because of that, I'd like to ship
> systemd unit definition files (and optimistically, getting them upstream).
> To do this, I need to settle on few defaults, which I'd like to discuss
> with
> you.
>
> 1) I'd like to propose running owserver as a central multiplexer;
> frontend software (owfs, owhttpd, owftpd) would ship with unit files
> pointing to and requiring owserver. Of course local admin can always
> override the default.
>
> That is the preferred arrangement. I would create a configuration file and
have all the programs share it (that allows consistent control of things
like temperature scale, bus master location, and even mount-point).
> Is there a problem with such arrangement?
>
> 2) For owfs, I'd like to encode default mountpoint in the unit file.
> What mount point is widest used? /mnt/owfs has ~80k mentions in
> google,
> /mnt/1wire about 600k. What's your preference?
>
> Completely arbitrary, but /mnt/1wire seems more obvious to me.
> 3) owfs: is --allow-other good option to have by default (with
> appropriate
> warning in unit file)?
>
Yes.You can then have a special user account running owfs and allow access.
>
> Any other things I should have in mind packaging owfs?
>
The one difficulty is knowing what bus-master to use. You can start owfs
based on seeing the USB bus-master. A serial adapter (or serial->usb)
bus-master can't be safely figured out on its own. i2c can safely scan for
bus-masters.
>
> Future work:
>
> I am clear that I'd like to see owfs package "just work" for majority of
> Fedora users. That means more use of facilities and automation provided
> by systemd. I know that primary development platform for OWFS is Debian,
> which
> is not systemd-only platform. So any patches would be rather hard to test
> by
> primary developers on Debian.
>
> I have a Fedora system for testing.
> Nevertheless, I image three step process in using systemd facilities.
> At each
> steps compatibility with other init systems is preserved, and only third
> step
> is (a little) intrusive for owfs.
>
>
> 1) ship unit files upstream; I think after some time of shipping and
> testing
> unit files in Fedora, they should be move to main owfs package. It's
> a matter
> of including couple text files + few autoconf snippets.
>
> 2) Right now I'm starting owfs (the FUSE thing) as normal service. I did
> not
> master using systemd's mount units with FUSE filesystems. When this
> is
> worked out, I envision shipping mount+autmount unit. This would make
> on-demand startup of owfs possible (and owserver as a dependency).
>
> 3) systemd provides facility for socket-activation; it's a mechanism
> comparable
> to beefed up xinetd. I'd like to patch owserver to have its socket
> managed
> by systemd. This would make few things possible:
> - removing explicit requirements for owserver from owhttpd and
> similar units
> - would allow owserver to be restarted without need to restart all
> frontend
> daemons.
>
> Any patches are welcome. I would like to avoid the difficulties of dynamic
library linking that plagues the Avahi code. Would the socket interface
break other platforms?
>
>
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_mar
_______________________________________________
Owfs-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/owfs-developers
