Divecomputer-reading from Mobile devices (Was: Divemate Fusion)

[Dirk Hohndel]

> On Apr 9, 2016, at 3:27 PM, Martin Gysel <me at bearsh.org> wrote:
> 
> Am 09.04.2016 um 23:55 schrieb Dirk Hohndel:
>> On Sat, Apr 09, 2016 at 02:46:32PM -0700, Linus Torvalds wrote:
>>> On Sat, Apr 9, 2016 at 1:59 PM, Dirk Hohndel <dirk at hohndel.org> wrote:
>>>> 
>>>> It would be equally easy to build a full Subsurface - we seem to have a
>>>> pretty complete Debian distro here.
>>> 
>>> Well, there's only 4GB of emmc flash (and 512MB RAM) on that thing, so
>>> I think it would be cramped to actually build subsurface on it.
>>> 
>>> But yeah, you don't actually want a UI, you really just want a process
>>> that links to libdivecomputer and can communicate over bluetooth to
>>> get commands and return results.
>> 
>> I have a simple BLE app but can't quite connect to it, yet. But I'm pretty
>> sure that this is just a matter of reading the docs one more time :-)
> 
> keep in mind, the throughput of a BLE connection is quite limited
> (~32kbps on iOS, maybe up to ~84kbps on Android [1]). to download more
> than a couple of dives a wifi connection might be the better option.
> furthermore already used technology like git could be used to transfer
> the dive from the 'device' to the mobile or to the cloud directly. BLE
> could be used to command it.

Yes, now that I have a working BLE connection from the C.H.I.P to my iPhone I would have to agree - it's not exactly fast.
But then again, the information we need to transfer isn't all that large, either. The compressed XML for me seems to average under 4kB per dive/divecomputer. That seems to imply that we could transfer a dive every couple of seconds or so. Definitely worth to play with. I'd much rather not mess with a wifi connection if I can avoid it. Since we can't assume that public wifi is available for the C.H.I.P to connect to we'd have to create a point to point wifi connection with the phone / tablet. Messy, complicated, error prone.

I don't know how much time I'll have to continue the proof of concept here, but I think this is looking surprisingly promising.

/D