[Container-tools] [atomic-devel] Nulecule Validator repo on GitHub

[Tomas Radej]

Hi.

How will releases be managed? If you put the validator into a directory 
in projectatomic/nulecule, you effectively can't tag commits as it will 
create a mess (tagging validator vs tagging all the rest of the repo). 
If you don't tag, how do you specify a hard point in time when release 
is made? I understand why you'd want to use e. g. a git commit hash for 
prerelease software, but the validator should be properly released once 
it's stable enough. Hunting down commits when debugging is a nightmare.

Regards, Tomas

On 20/05/15 09:57, Václav Pavlín wrote:
> Hi,
>
> I'd like to see this living inside the spec repo. As long as it is only
> able to validate Nulecule files and has "nulecule" in the name, I don't
> think it makes much sense to maintain it separately. I agree with you it
> does not make sense to merge it with atomicapp although we should figure
> out a way how to utilize it there.
>
> I am also CC'ing container-tools ML where most of the nulecule related
> discussion happens.
>
> Regards,
> Vašek
>
> On 19.5.2015 17:23, Tomas Radej wrote:
>> Hi,
>>
>> I have made a validator library/executable for Nulecule files [1], and
>> I'd like to transfer it to the Project Atomic organisation on GitHub.
>> There has been some discussion about where it should go, if it should
>> be merged with the main spec repo [2] or the atomicapp [3], or
>> something completely different. Mind you that this lib is just a
>> syntactic validator, not a linter, so it has can't understand the
>> files beyond simple checking against a schema.
>>
>> I think that the nulecule_validator repo should stay as a stand-alone
>> repository with its own releases and lifecycle. The reason is that it
>> can be easily used on its own. It only needs the schemata from the
>> main spec repo, nothing else, and these schemata can be linked to it
>> via git submodule or subtree.
>>
>> If it was merged with atomicapp, every user wanting to validate their
>> file on any machine would need all dependencies of the atomicapp
>> installed (Docker?). A typical case would be a user developing a
>> Nulecule app with DevAssistant on one (bare metal) machine and
>> deploying on another (virtual one?). In addition to that, the
>> validator wouldn't be much use to alternative implementations of the
>> Nulecule spec either.
>>
>> Questions, comments, suggestions?
>>
>> Thanks, Tomas Radej
>>
>> [1] https://github.com/tradej/nulecule_validator
>> [2] https://github.com/projectatomic/nulecule
>> [3] https://github.com/projectatomic/atomicapp
>>
>