Re: [PATCH v4 1/5] implement isar-bootstrap using debootstrap

[Alexander Smirnov <asmirnov@ilbers.de>]

On 03/15/2018 11:05 AM, Claudius Heine wrote:
> Hi Alex, hi Jan,
> 
> On 03/14/2018 07:53 PM, Jan Kiszka wrote:
>> On 2018-03-14 07:25, Alexander Smirnov wrote:
>>> On 03/14/2018 05:14 PM, Claudius Heine wrote:
>>>> Hi Alex,
>>>>
>>>> On Wed, 2018-03-14 at 13:38 +0300, Alexander Smirnov wrote:
>>>>> On 03/14/2018 12:52 PM, Claudius Heine wrote:
>>>>>> Hi Alex,
>>>>>>
>>>>>> On Wed, 2018-03-14 at 11:56 +0300, Alexander Smirnov wrote:
>>>>>>> On 03/07/2018 07:44 PM, claudius.heine.ext@siemens.com wrote:
>>>>>>>> From: Claudius Heine <ch@denx.de>
>>>>>>>>
>>>>>>>> Since multistrap is deprecated for some years, it is required
>>>>>>>> to
>>>>>>>> change
>>>>>>>> to debootstrap.
>>>>>>>>
>>>>>>>> This patch introduces the 'isar-bootstrap' recipe that
>>>>>>>> implement
>>>>>>>> the
>>>>>>>> creation of a minimal base root file system using debootstrap.
>>>>>>>>
>>>>>>>
>>>>>>> Why do you use term 'bootstrap'? It's completely different topic
>>>>>>> than
>>>>>>> debootstrap.
>>>>>>
>>>>>> Because I want to be implementation independent. If for some reason
>>>>>> someone wants to change the name of the tool again (I know its not
>>>>>> to
>>>>>> be expected), this way only the implementation of the isar-
>>>>>> bootstrap
>>>>>> needs to be changed.
>>>>>> In theory with an alternative implementation of some interfaces
>>>>>> other
>>>>>> distribution could be supported. I don't know in which direction
>>>>>> isar
>>>>>> will go, but binding your interface names to tightly to the name of
>>>>>> some third-party products is not a good idea.
>>>>>>
>>>>>> What I tried is called loose coupling and that is generally seen as
>>>>>> a
>>>>>> good software design pattern.
>>>>>>
>>>>>>>     According to the wiki:
>>>>>>>
>>>>>>> https://wiki.debian.org/DebianBootstrap#Bootstrapping
>>>>>>>
>>>>>>> Bootstraping is a process of creation Debian from zero to a full
>>>>>>> archive.
>>>>>>
>>>>>> The word "bootstrapping" is used in many projects and generally
>>>>>> describes starting something from almost nothing. Also this
>>>>>> patchset
>>>>>> doesn't use "DebianBootstrap" is uses "isar-bootstrap".
>>>>>
>>>>> In this case just do not use terms from Debian world.
>>>>
>>>> I didn't. I use 'isar-bootstrap' and neither it nor its parts 'isar'
>>>> nor 'bootstrap' is a term that is exclusively used with Debian.
>>>>
>>>> 'bootstrap' is a common term in computer science with many different
>>>> meaning. If someone heard 'bootstrap' only used by the Debian project
>>>> before, that that is their problem. I think I heard it first in some
>>>> projects as the 'bootstrap.sh' shell script that takes care about
>>>> initializing the build environment correctly. I might have been a bit
>>>> confused when I heard that they also describe the first compilation of
>>>> compiler by an other one written in a different language as
>>>> bootstrapping, but that might just be because English is my second
>>>> language.
>>>>
>>>>>    It would be enough
>>>>> to name it like 'base-rootfs'. That's for example, how 'buildchroot'
>>>>> term was invented, to avoid intersections with 'sysroot' and
>>>>> 'buildroot'.
>>>>
>>>> 'sysroot' and 'buildroot' are names that where invented by someone,
>>>> 'bootstrap' is a word, like 'face' or 'book'. I don't automatically
>>>> think of 'facebook' just because I read the word 'book' or 'face' in
>>>> some other context.
>>>>
>>>> Personally I don't like to add another '*base*' thing to isar. We
>>>
>>> I didn't ask to use it, I only tried to describe the idea. Whatever else
>>> could be used, but without intersections with Debian.
> 
> So then why did you bring this up if you don't have an alternative 
> suggestion I should use instead? That just sounds like destructive 
> criticism.
> 
> Since I don't think that there is an intersection in the nomenclatures 
> of Debian with the name 'isar-bootstrap' I choose to use this name.
> 
> Or should I now start guessing what other name you might prefer? What is 
> this?
> 
>>>
>>>> already have 'isar-image-base' and 'dpkg-base'. And this recipe is
>>>> should not be used as a 'base' to be included or inherited upon like
>>>> the others. So 'base-roofs' as a recipe name does not fit IMO.
>>>>
>>>> (Also when I hear 'base' I think about numbers and get very confused ;)
>>>>
>>>> I still prefer 'isar-bootstrap' since its output should be what of some
>>>>    kind of distro-specific system bootstrap machanism, like 
>>>> debootstrap,
>>>> febootstrap (now supermin), pacstrap, ..., generates.
>>>> If the distro-specific bootstrap mechanism builds its 'bootstraped root
>>>> file system' from a bunch of binary packages or directly from source is
>>>> implementation and distro specific. So the distinction only makes sense
>>>> in the upstream project but not here, since we are just using what the
>>>> upstream distributions provides for general consumption.
>>>
>>> I asked the question, but you cut it, so I'm going to repost the whole
>>> quote here:
>>>
>>> 8<--
>>>
>>> AFAIK there are 2 kinds opinions in Debian community about bootstraping:
>>>   - Build from sources (DebianBootstrap, rebootstrap)
>>>   - Install from debs (debootstrap)
>>>
>>> My question, why it's so important to use initially ambiguous term,
>>> which may lead to potential confusing for Debian users?
>>>
>>> 8<--
>>
>> I hope we can settle on this topic quickly, specifically as this
>> internal naming of classes are not really user-facing.
> 

It's not only about internal class names, this series introduces 
'do_bootstrap' task that is visible to user and will be a part of 
documentation.

My position is simple: I read patches and if I've found something 
unclear, I'm trying to understand what was the reason going this way.

Regarding bootstraping, I provided above 2 kinds of interpretation of 
this process in Debian, so this term is initially ambiguous.

If you use deboostrap, it's ok to name tasks/recipes/etc respectively 
(do_debootstrap, ...). If in future this code will be refactored to 
support other rootfs creation mechanisms, it'll be very easy to rename 
these files/classes. Moreover, at the moment there is no other option 
for debootstrap, so it's only about something hypothetical in future, 
what could never happen.

@Jan:
I've got your opinion, so nothing else from me.

> I really hope so as well. Arguing about internal names is very annoying.
> 

@Claudius:
For me is very annoying when my question is ignored after explicit 
re-posting. So let's be polite and keep conversations here free of emotions.

Alex