>    On 2 December 2013 13:09, Aliaksei Katovich <[1][email protected]>
>    wrote:
> 
>              * complex solution (reconstruction) vs simple
>      (thin-provisioning);
>              * if I understand you correctly reconstruction implies
>      package
>                re-installation which might become a problem due to:
>              - time consumed;
>              - space consumed;
>              - dependencies;
>              * how to rollback in case reconstructed configuration
>      failed: files on
>              system got already replaced with new ones;
> 
>    You just expect the device to be synchronized with the cloud, so that
>    there is nothing unique
>    on the device that can be lost.

        I still do not see how cloud can save someones troubles with faulty
        update if update breaks connectivity.

>    Again, this has some requirements, but afaik all the current
>    Android/iOS phones
>    have the same expectations, so it's not something unthinkable,
>    industry-wise.
> 
>              * how to recover from update if connectivity framework gets
>      broken right
>                after.
> 
>    You could have a non-upgradable partition that has working everything,
>    albeit in a minimalistic way.

        This is called restore point either made at once during boot or
        another time during previous update. Restore point is based on
        thin-provisioning and there is no need to waste space for extra
        partitions.

>    This is how MACs do it.
> 
>              IMO snapshot based rollback is cheaper to make and maintain
>      than
>              rebuilding.
> 
>    Well, but you cannot really choose, right?

        Sorry, did not get it, choose what?

>    If the device is lost/stolen, there is _nothing_ to roll back.
>    Unless you want to upload to the cloud hte whole snapshot, but that
>    would be
>    even worse, because it would require very large storage per user, while
>    downloading
>    from some repo can be scaled gracefully, even dynamically.
>    So reconstruction must be supported anyway and it must function well.
>    At that point, why not leveraging it also for this purpose?

        Ah, now I seem to get your point.

        If my understanding is correct, then we are talking about two different
        things handling two different use-cases:

        1) rollbacks in case devices fails to act stable or at all (my proposal)

        vs.

        2) replication of device configuration if original device is 
lost/stolen.

        There is no doubt that these features can co-exist.

--
Aliaksei

>    --
>    cheers, igor
> 
> References
> 
>    1. mailto:[email protected]
_______________________________________________
Dev mailing list
[email protected]
https://lists.tizen.org/listinfo/dev

Reply via email to