I tend to agree with Arek on this one.  It feels a good logical division to
have it separate, primarily so when updates to the Azure SDK are released,
a new version of this module can be released without having to update
oak-blob-cloud.  By that same argument it may make sense at some point to
do the same thing for AWS and S3DataStore.

-MR

On Wed, Mar 15, 2017 at 3:55 AM, Arek Kita <[email protected]> wrote:

> Hi Amit,
>
> On 15/03/2017, 10:29, "[email protected] on behalf of Amit Jain" <
> [email protected] on behalf of [email protected]> wrote:
>
> > Hi Team,
> >
> > There is a new contribution for azure blob storage support - OAK-4933.
> > This introduces a new module oak-blob-cloud-azure. This certainly seems
> to
> > be the right approach from a separation and deployment standpoint. But in
> > terms of code the module may only contain a few classes and also in
> future
> > if we introduce (on my horizon) support for jclouds we could unify all
> > under one module.
> >
> > What is the correct thing to do here?
>
> I IMHO think that even if it contains a few classes, there are/there will
> be for sure some external transitive dependencies that could be helpful to
> have in independent deployable module. I think it should not be about how
> many classes but rather how mature the code is and how frequently it will
> be updated or backported (code maturity level).
>
> I would go with a separate module for this so you can replace it, leaving
> the rest as is (possibly).
>
> Regrads,
> Arek
>
>

Reply via email to