Images and cloud storage

[K. "pestophagous" Heller]

On Thu, Dec 3, 2015 at 10:53 PM, Dirk Hohndel <dirk at hohndel.org> wrote:
> On Fri, Dec 04, 2015 at 08:47:07AM +0200, Willem Ferguson wrote:
>> Question. The option of "synch dive log in background" is activated and I
>> save dives to the Subsurface cloud with the images on a non-connected source
>> (so the images are referenced in the log but they are not accessible). If,
>> after saving to cloud, I now connect the drive with the images to view these
>> images within Subsurface, would the background process see these images and
>> synch them?
>
> I don't think so.
> But fundamentally we need to redesign this.
>
>> About thumbnails. I think it would be super if it were possible to store
>> thumbnails of the images in the repo. This would allow cursory inspection of
>> the images associated with a dive, even if the full-size images are only
>> available on an external drive that is not connected at that specific point
>> in time. But even this is not an easy issue. I have been playing around with
>> scaled down jpeg versions of some of my images and, for having a usable
>> thumbnail, I cannot get the size of a 200 by 150 pixel image below about 3
>> kB. If one has 1000 users each with 10000 images, the disk space required
>> would still not be trivial at all and, for trigger-happy photographers, the
>> thumbnails may easily need more space than the dive log itself. And in real
>> life there may be more than 1000 users with images using the cloud storage.
>> And then, from a software point of view, the thumbnails need to be created,
>> stored and correctly referenced by Subsurface.
>
> I'm not too worried about the storage (yeah right - whatever - I'm paying
> so much for this hobby, what's a few extra TB of cloud data I pay for...).
> The thumbnail creation really is easy. I just don't know if this is what I
> want.
>
> I'm thinking of going back to just not storing images in git, period.
>
> /D

(some of what follows may consist of "dumb questions." i have never
looked at the subsurface source code that interacts with git, although
now I am feeling very curious to do so.)

it makes sense to me that disk storage is not a big concern, but
rather that the amount of data transferred over slow phone/tablet
connection is a concern.

Question: if the pictures to be transferred were small thumbnails,
then wouldn't we solve the data-transfer problem by doing something
like "shallow clone" when downloading the pictures to a phone? (if
each picture is very small, then the only problem with transferring
those small git-tracked items would be if there was a lot of history,
right?)

we definitely need to sync pictures (or thumbnails) somehow. it's a
great feature.

Bad ideas: other (bad) things coming to mind are git-annex and git LFS
(?). but those don't seem like fun things to have a dependency on.

lastly, this might sound the most heretical of all, and yet it might
be simplest:
     what about (for the pictures only) using SVN? at least it would
be a well-understood, stable C library (not haskell or ruby like
git-annex and git LFS).

     librsync?

ok. enough mind-dump of ill-considered ideas for now :)
/K