Re: Filesystem Hierarchy Standard (FHS) and NetBSD

[Adam Hoka <adam.hoka%gmail.com@localhost>]

On Tue, 10 May 2011 12:36:06 +0200
Julian Fagir <gnrp%komkon2.de@localhost> wrote:

> Hi,
> 
> > > Given that it describes a particular file hierarchy conventionally
> > > used only by Linux systems from major Linux distributors, how would
> > > you expect that to be the case?
> > 
> >  From the opening paragraph of the standard:
> > 
> > "This standard consists of a set of requirements and guidelines for file 
> > and directory placement under UNIX-like operating systems. The 
> > guidelines are intended to support interoperability of applications, 
> > system administration tools, development tools, and scripts as well as 
> > greater uniformity of documentation for these systems."
> > 
> > My goal is to determine if this is actually true, or if it is only 
> > relevant to Linux, as you suggest.  I don't want to exclude you (and the 
> > other BSD systems, and the other UNIXes) without making sure.
> far from being away a *BSD- or Linux-developer, my opinion:
> I don't think you can reach this uniformity without having only a very small
> standard.
> 
> Adding BSDs to the FHS would need various modifications which just do not
> apply to Linux-systems. Just, for example:
>  * /var/db is rarely used on Linuxes, on BSDs the packaging informations are
>    stored there, /var/lib is rather used on Linux.
>  * /boot is not the place to store boot-files, the configuration and kernels
>    on NetBSD lie partially in /.

But we could use that as the current place for these files are just keeping
the status quo.

>  * /rescue for the rescue-binaries is not even mentioned in the FHS, as
>    is /exports for NFS-exports, which is not a standard, but often used.
>  * /var/cron is specified as 'reserved' but historical, though used on BSDs
>    for cron.
> 
> Thus, even the directories in the standard and not put into the appendix are
> platform-specific and not applicable to all.
> 
> So, when trying to be a standard for *all* Unixes, there would be three ways:
>  1. Be as generic as possible and include only minimal specifications.
>     This would be an immutable standard, as the only generics of all Unixes
>     are historical. This is not what the FHS is intended for and would rather
>     suit on a hobbyist page.
>  2. Include all flavours for all Unixes and make a big standard.
>     Taking this way, you would have plenty of directories considered 'used
>     only on system X, Y and Z', making the standard large and unreadable if
>     you want information about a specific distribution, thus making the
>     standard rather informational than a specification.
>  3. Find a way between 1. and 2., considering only the majororities.
>     When do you consider a Unix big enough to be part of the standard? This is
>     a NetBSD-mailing list, which has a usage of <10% considering only BSDs,
>     now think about DragonFly.
> 
> Additionally, all Unixes except Linux have a more closed development with
> only one community to consider, making it easier to introduce new
> hierarchies, whilst the term Linux incorporates hundreds of distributions and
> flavours how to do things.
> 
> This is just my view as a BSD- and Linux-user, not even thinking of Solaris
> and all other active Unix-derivatives.
> 
> Therefore, I think, making the FHS a standard for all Unixes is not possible
> nor wanted.
> What I would rather think of would be explicit mentioning that the FHS as
> currently published is for Linuxes ('LFHS'), merging Chapter 6 into the
> standard itself. Perhaps, one could extract an informal smaller standard
> which is applicable to more systems, but as I said, this would rather fit on
> a hobbyist page than on a standard which is meant to be stating a standard,
> not informing about it.
> 
> 
> Regards, Julian


-- 
NetBSD - Simplicity is prerequisite for reliability