From 1be510f9529cb082f802408b472a77d074b394c0 Mon Sep 17 00:00:00 2001 From: Nicolas Vigier Date: Sun, 14 Apr 2013 13:46:12 +0000 Subject: Add zarb MLs html archives --- zarb-ml/mageia-dev/2012-September/018674.html | 135 ++++++++++++++++++++++++++ 1 file changed, 135 insertions(+) create mode 100644 zarb-ml/mageia-dev/2012-September/018674.html (limited to 'zarb-ml/mageia-dev/2012-September/018674.html') diff --git a/zarb-ml/mageia-dev/2012-September/018674.html b/zarb-ml/mageia-dev/2012-September/018674.html new file mode 100644 index 000000000..e64ce310e --- /dev/null +++ b/zarb-ml/mageia-dev/2012-September/018674.html @@ -0,0 +1,135 @@ + + + + [Mageia-dev] /run vs /var/run in configuration files + + + + + + + + + +

[Mageia-dev] /run vs /var/run in configuration files

+ Guillaume Rousse + guillomovitch at gmail.com +
+ Fri Sep 14 15:24:31 CEST 2012 +

+
+ +
Le 11/09/2012 13:44, Colin Guthrie a écrit :
+> 'Twas brillig, and Guillaume Rousse at 11/09/12 11:47 did gyre and gimble:
+>> Le 11/09/2012 12:23, Sander Lepik a écrit :
+>>> 11.09.2012 13:15, Guillaume Rousse kirjutas:
+>>>> [...]
+>>>>
+>>>> The /run - /var/run merge (/usrmove) is supposed to make the change
+>>>> transparent for applications. Manually converting applications to
+>>>> explicitely refers to the new location doesn't change its usefulness.
+>>>>
+>>> Well, if your system can't mount /var for some reason then keeping
+>>> things on /run might make recovery easier.
+>
+>> Then we probably should not refer to systemd directory as
+>> /usr/lib/systemd, but as /lib/systemd for exactly the same reason.
+>
+> I don't really see the comparison here.
+The point was: if the canonical path for /usr/lib|/lib directory (which 
+is actually the same) if the longuest one, why should the canonical path 
+for the /var/run|/run directory be the shortest one ?
+
+> systemd is build to use /usr/lib/systemd these days, so it's technically
+> incorrect to refer to it as anything else (of course technically
+> incorrect != practically incorrect!)
+>
+>> Also, the usefulness of an available /run when /var is quite discutable,
+>> whereas its main purpose is to store pid files for backgroud processes,
+>> most of them using /var/lib for storing their data.
+>
+> Well, not just pid files. It also stores socket files which are critical
+> for IPC and such like and is an important bridge between initrd and the
+> real rootfs.
+>
+> Personally I'm more in favour of using /run directly and adding lint
+> rules that cause a build failure if files are packaged in either
+> /var/run or /run.
+Shipping /var/run or /run files/directory is a different issues, and is 
+rather related to tmpfs conversion.
+
+However, a distinct lint rule could be to disallow the creation of 
+/var/run subdirectories in tmpfiles.d configuration files, or to use 
+/var/run pathes for PID files in systemd unit configuration files, if 
+/run is to be preferred.
+
+> If nothing else it's marginally more efficient to use /run. No spinning
+> media stat required to dereference the index. In extreme cases, it will
+> help extend battery life. The benefits are tiny I'm sure, but it does
+> depend on context/setup and if there is no other tangible reasons, maybe
+> even this small advantage is enough to sway in one direction rather than
+> the other!
+Well, I'm not really impressed by potential perf issues, especially with 
+no backing benchmarks. I'm rather concerned about ease of maintainance, 
+and consistency between spec files.
+
+Anyway, I just found out than /usr/lib and /run are the actual 
+directory, with /lib and /var/run the symlinks. Which makes an argument 
+of favor of considering the first one as canonical.
+
+-- 
+BOFH excuse #182:
+
+endothermal recalibration
+
+ + + + + + + + + + + + + + + + + + + + +
+

+ +
+More information about the Mageia-dev +mailing list
+ -- cgit v1.2.1