[oe] base-files vs. FHS-2.3

[Steffen Sledz]

Am 10.02.2011 15:32, schrieb Mark Hatle:
> On 2/10/11 6:03 AM, Phil Blundell wrote:
>> On Thu, 2011-02-10 at 11:55 +0100, Steffen Sledz wrote:
>>> While dealing with the /var subdirs i hit some differences between Filesystem Hierarchy Standard (current FHS-2.3) and the base-files package.
>>>
>>> One point is that /var/tmp is symlinked to volatile/tmp but according to the standard this dir contains "Temporary files *preserved between system reboots* ".
>>>
>>> Also /var/cache is symlinked to volatile/cache but the standard describes "The data must remain valid between invocations of the application *and rebooting the system.* " what i would read in the way that /var/cache should be persistent too.
>>>
>>> May be there are other differences.
>>>
>>> In my opinion the symlinking in base-files is buggy and should be fixed. Or am i wrong here?
>>
>> It probably would be good to have a version of base-files which was
>> fully FHS conformant.  But there are a significant number of OE target
>> systems where it is simply impossible to comply with these requirements
>> since there is no persistent read/write storage available: the only
>> choice is between flash (persistent but read-only) and ramdisk
>> (read-write but volatile).  
>>
>> Clearly, placing /var/tmp or /var/cache in a readonly location is
>> unlikely to produce any useful results so linking them into volatile/ is
>> the least bad option in that situation.  Any change to base-files would
>> need to be done with some level of care in order to not break those
>> kinds of setups which do work today.  I guess it should be a DISTRO
>> decision whether or not to adhere to the FHS in this area.
> 
> I echo this.  In my experiences the default filesystem should be FHS compliant,
> but there must be a variant that works for specific embedded projects that
> simply do not have writable storage.
> 
> The ramifications of this are that /var/tmp, /var/cache are both transient over
> reboots.  Almost all applications will deal with this gracefully, so it's not
> really an issue.  It is something to be aware of though when doing software
> integration and validation that the problems must be able to create tmp and
> cache files over again if the data is not already on the system.
> 
> The important things when following the FHS really are R/O vs R/W support, as
> well a boot vs run-time.  (i.e. things required on '/' vs things that should
> live in '/usr'...)

Ping!

-- 
DResearch Fahrzeugelektronik GmbH
Otto-Schmirgal-Str. 3, 10319 Berlin, Germany
Tel: +49 30 515932-237 mailto:sledz at DResearch.de
Fax: +49 30 515932-299
Geschäftsführer: Dr. Michael Weber, Werner Mögle;
Amtsgericht Berlin Charlottenburg; HRB 130120 B;
Ust.-IDNr. DE273952058