Precedence: list
MIME-Version: 1.0
In-Reply-To: <B4C204D8-C445-4347-9DDA-FFA89CB8A287@jirutka.cz>
References: <c0880cc02818948a5b400087209c80d0@riseup.net> <AA21151C-A9BB-4FD9-9199-0F57434B8047@jirutka.cz>
 <174bbc5214754cfa31d9156ec1e78660@riseup.net> <CAJDAfTD9+q3fbYMJp05sLLhHVSdKCzNU-diEtjCyzKXUED4iiA@mail.gmail.com>
 <B4C204D8-C445-4347-9DDA-FFA89CB8A287@jirutka.cz>
From: Alba Pompeo <albapompeo@gmail.com>
Date: Sun, 3 Jul 2016 11:36:47 -0300
Message-ID: <CAJDAfTDJbu7GiQxcw0a-o9_3nWMJk_woW51mqBzmppp1YyQFew@mail.gmail.com>
Subject: Re: [alpine-devel] smdev
To: Jakub Jirutka <jakub@jirutka.cz>
Cc: =?UTF-8?B?Q8OhZw==?= <caoc@riseup.net>, alpine-devel@lists.alpinelinux.org
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

> Busybox is a base of the Alpine=E2=80=99s userland, so we have it anyway.

ncopa said he would welcome sbase [0] and ubase [1] once they are
stable, because of its advantages (such as correct handling of UTF-8).
For example, try the following on Busybox and see how it fails. echo
"=C3=B5=C3=A3" | tr -d "=C3=B5". On sbase it works.

> (e)udev=E2=80=99s monster complexity is needed maybe just for desktops if=
 at all (not sure, I don=E2=80=99t use Linux on desktop).

That's why Alpine hasn't adopted smdev yet, because there are problems
to be fixed (like hotplugging, bluetooth, etc). I'm not sure the exact
issues.

> Yes, but just until you need to change something=E2=80=A6
> Maybe I=E2=80=99m biased here because I=E2=80=99ve already needed to modi=
fy the Alpine=E2=80=99s mdev.conf several times.

Most users rarely if ever need to change something in such low-level
system programs.

> This means to recompile the smdev binary

Will people really need to recompile it? Alpine should provide a
default config.h that works for 99.9% of users.

> We need BusyBox anyway

Not forever.

> smdev package would not replace mdev for all use cases

What's missing on smdev? Developers said in the past they want to help
with anything Alpine needs.

> imagine how you tell average user to recompile smdev himself/herself inst=
ead of just modify simple config file

The average user doesn't know what's mdev/smdev/etc and won't need to
change the Alpine sane defaults. That's how I see it...


[0] http://git.suckless.org/sbase
[1] http://git.suckless.org/ubase

On Sun, Jul 3, 2016 at 11:13 AM, Jakub Jirutka <jakub@jirutka.cz> wrote:
>>> =E2=80=A6it doesn=E2=80=99t have all its features.
>
> What features does it miss exactly? I can=E2=80=99t find in on the provid=
ed page or inside the smdev=E2=80=99s repository.
>
>>> The benefits are: you don=E2=80=99t need BusyBox for it=E2=80=A6
>
> Busybox is a base of the Alpine=E2=80=99s userland, so we have it anyway.
>
>>> By the way, I don=E2=80=99t know why Alpine doesn't use mdev, to me it =
seems to be the most appropriate variant=E2=80=A6
>
> Alpine does use mdev by default (at least in alpine-virt variant, not sur=
e about full-fledged).
>
> I personally use mdev on all my servers, VMs and bear metal, Alpine and G=
entoo. I believe that mdev is fully sufficient for most of the server and e=
mbedded use cases. (e)udev=E2=80=99s monster complexity is needed maybe jus=
t for desktops if at all (not sure, I don=E2=80=99t use Linux on desktop).
>
>> I agree with C=C3=A1g about the absence of textual configuration files. =
It=E2=80=99s an advantage because it keeps complexity down.
>
> Yes, but just until you need to change something=E2=80=A6
>
>> Almost no user will change their smdev config options=E2=80=A6
>
> Maybe I=E2=80=99m biased here because I=E2=80=99ve already needed to modi=
fy the Alpine=E2=80=99s mdev.conf several times.
>
>> But if the user really wants to, I see no problem with config.def.h
>
> This means to recompile the smdev binary (note: you need gcc). Then you c=
an install it directly from the sources and it=E2=80=99ll be the same, i.e.=
 you will not benefit from aport. Overwriting binaries managed by apk is re=
ally not a good practice IMO.
>
>
> I really don=E2=80=99t see any *practical* benefit for Alpine. We need Bu=
syBox anyway, so why to replace BusyBox=E2=80=99s mdev with smdev that is j=
ust less flexible (no config file)? Also it will mean supporting three inst=
ead of two (mdev, eudev) udev packages, because smdev package would not rep=
lace mdev for all use cases (imagine how you tell average user to recompile=
 smdev himself/herself instead of just modify simple config file; Alpine is=
 not LFS or Gentoo).
>
> Jakub
>
>
> On 3. Jul 2016, at 15:51, Alba Pompeo <albapompeo@gmail.com> wrote:
>
>> I agree with C=C3=A1g about the absence of textual configuration files.
>> It's an advantage because it keeps complexity down.
>> Almost no user will change their smdev config options, that's
>> something Alpine will have sane defaults for.
>> But if the user really wants to, I see no problem with config.def.h
>> [0]. /usr/src to the rescue.
>> I remember smdev developers talked about wanting to integrate it on
>> Alpine in the past, here on the mailing list.
>> I think ncopa wants it too, but had problems with hotplugging (?) I thin=
k.
>> But it looks like C=C3=A1g has solved that problem. Let's wait and see i=
f
>> ncopa thinks now smdev is ready (for 3.5 alpine release?).
>>
>> Ciao!
>>
>> [0] http://git.suckless.org/smdev/tree/config.def.h
>>
>> On Sun, Jul 3, 2016 at 10:35 AM, C=C3=A1g <caoc@riseup.net> wrote:
>>> Jakub Jirutka wrote:
>>>
>>>> I wonder, what are the differences between mdev and smdev? What are
>>>> the benefits of smdev for end users?
>>>
>>>
>>> smdev is IIRC a fork of mdev (or maybe was, probably it was rewritten)
>>> that is compatible with it [0], though it doesn't have all its features=
.
>>> The benefits are: you don't need BusyBox for it and it is smaller
>>> (almost thrice as short in terms of LOC). It does what it says -
>>> manages device nodes, and does it well.
>>>
>>>> I currently see one significant drawback - smdev (and all other
>>>> suckless projects) doesn't support textual configuration files. This
>>>> approach doesn't go well with binary distributions; users can't simply
>>>> change any settings. I understand the reasons behind using header
>>>> files for config, but the fact is that it's not very convenient for
>>>> users.
>>>
>>>
>>> For me, it's rather an advantage. Once configured you go with it. Actua=
lly
>>> I've never modified my udev rules, they always worked for me and with s=
mdev
>>> I use
>>> the default config.h because it works just fine. Anyway, if users need =
to
>>> modify the rules, I'd put the source inside /usr/src and let them do it=
 :)
>>>
>>> By the way, I don't know why Alpine doesn't use mdev, to me it seems to=
 be
>>> the
>>> most appropriate variant (or does it? I heard it uses eudev. I currentl=
y
>>> have
>>> no chance to check it out).
>>>
>>> If ye need more info about smdev, please consult the link down below an=
d the
>>> code.
>>> Also, write to dev at suckless dot org, since I am not an smdev develop=
er.
>>>
>>> Sl=C3=A1inte,
>>> C=C3=A1g
>>>
>>> [0]:http://core.suckless.org/smdev
>>>
>>>
>>>
>>> ---
>>> Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
>>> Help:         alpine-devel+help@lists.alpinelinux.org
>>> ---
>>>
>>
>>
>> ---
>> Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
>> Help:         alpine-devel+help@lists.alpinelinux.org
>> ---
>>
>


---
Unsubscribe:  alpine-devel+unsubscribe@lists.alpinelinux.org
Help:         alpine-devel+help@lists.alpinelinux.org
---