bug#29932: [PATCH 0/2] Clean up operating-system-kernel-arguments.

[Maxim Cournoyer <maxim.cournoyer@gmail.com>]

Hello,

Maxim Cournoyer <maxim.cournoyer@gmail.com> writes:

> Hello,
>
> Danny Milosavljevic <dannym@scratchpost.org> writes:
>
>> Hi Ludo,
>>
>>> I’m a bit lost: in my tree I don’t have
>>> ‘operating-system-boot-kernel-arguments’.  Is it still pending?
>>
>> It's added by PATCH v2 1/2 from the series.  Didn't the second mail get through?
>>
>>> Otherwise my only question is whether it’s a good idea to move away from
>>> the ‘user-’ convention.  On one hand, it’s the convention we also have
>>> for services (‘-user-services’ vs. ‘-services’), so it would be a good
>>> thing to remain consistent.  OTOH, what you propose is maybe clearer.
>>> 
>>> Thoughts?
>>
>> Yeah, I've split it into two patches because I actually got used to
>> operating-system-user-kernel-arguments by now (only a few days in).
>> We could only apply PATCH v2 1/2 and not apply PATCH v2 2/2 if we
>> wanted.
>>
>> In the end it comes down to whether we deem the existence
>> operating-system-boot-kernel-arguments an implementation detail or not
>> (whether the user would ever need to be aware of
>> operating-system-boot-kernel-arguments).  We have to export
>> operating-system-boot-kernel-arguments because one thing in
>> gnu/system/vm.scm needs it - otherwise it would be very much an
>> implementation detail.
>>
>> Let's see what the others say.
>
> Two years later, here's what I have to say :-)
>
> I think it's nice, as a user, to be able to inspect the dynamically
> computed kernel arguments that Guix would use, as that can be used for
> debugging and gaining a better understanding (e.g., when passing an
> argument option that overrides one computed by Guix).
>
> If I followed this discussion correctly, currently we have:
>
> 1. operating-system-kernel-arguments which is a combination of
> dynamically computed arguments by Guix + the users arguments and
> 2. operating-system-user-arguments which are the users arguments
> themselves.
>
> It is proposed here to split this into:
>
> 1. operating-system-boot-kernel-arguments for the Guix-computed ones
> 2. operating-system-user-kernel-arguments remains unchanged
>
> Thus if the user wants to know what boot arguments their system will
> use, they'd have to append these two together.
>
> I think that two years have elapsed without touching this is perhaps an
> indication that it doesn't address any real problem :-).  While it's
> good to attempt to clarify things, I'm afraid that changing this would
> confuse more that it'd help.  As Ludovic pointed out, it'd also clash
> with the convention currently in use for services.
>
> What you do think?

There haven't been any further comments.

Closing.

Maxim