[OE-core] ROOT_HOME: /home/root

[Laszlo Papp]

On Wed, Jan 29, 2014 at 1:13 PM, Richard Purdie
<richard.purdie at linuxfoundation.org> wrote:
> On Wed, 2014-01-29 at 12:59 +0000, Laszlo Papp wrote:
>> On Wed, Jan 29, 2014 at 12:52 PM, Richard Purdie
>> <richard.purdie at linuxfoundation.org> wrote:
>> > On Wed, 2014-01-29 at 12:32 +0000, Laszlo Papp wrote:
>
>> > By having a default of /home/root/ we can catch software that has issues
>> > with relocation of that.
>>
>> I am not sure what you mean. Could you please elaborate?
>
> As you mention, "/root" is more standard. It therefore becomes hard to
> spot software that assumes this rather than using the directory we
> configure it to if the default is also /root.
>
> By having a slightly more unusual default choice, we quickly find the
> software that doesn't adapt to our variable.

You seem to be referring to testing rather than forcing a test case
for the software, and hence majority. There are two better
alternatives IMHO to address this:

a) QA team defines such a test case.

b) In addition to a), you would get reports from users who set it to
/home/root if there are issues. Although, I am certainly questioning
that if more than 10-20% was setting it to "/home/root" at all, or
even much less.

Daniel in the CC also wrote an issue on IRC that it could lead to a
non-loginable system if the /home does not get mounted which would
better not be mandatory without setting it explicitly IMHO.

>> > Having the writeable user data in one directory like this is useful
>> for several classes of embedded style devices.
>>
>> Could you please provide any examples?
>
> One that springs to mind is the Sharp Zaurus series of PDAs have
> separate /home partitions in flash. You can reflash a new rootfs without
> overwriting the user config data.

I addressed this issue in my initial email, but I will reiterate this
concern with two solutions:

a) You can do package update rather than reflashing.

b) Even if a) does not work, as written in my initial email, you can
have a /root partition.

Still, the fact that you can only mention one use here, is somewhat
also self-explanatory that it is not the majority.

>> > So to be honest I don't see a pressing reason to change this.
>>
>> I do, because the earlier it is done, the fewer users that may have
>> incompatible changes. As the time goes ahead, more and more users will
>> stick to it as "default". I believe this means those who do not care
>> about proper Unix separation.
>
> Its been like this for years and seems to work perfectly fine for
> people.

I do not see it perfectly fine that if the majority of the people get
the extra work rather than the minority. I saw this several times now
in the history that it was done manually on the end-users side, and
since it is not just me, I thought I would bring this topic up.

I would rather prefer to ease the end users' work wherever it is possible.

> You can configure it just fine. As I said previously, I see no
> pressing reason to change it.

As written before, you could configure the other way around, too. The
question boils down to that: who should configure it on its own?

* The person who tries to follow the unix philosophy and the generic
behavior on most of the Unix systems out there?

* The person who follows some rare use case?

My opinion is the former.