Re: [Fwd: local device discovery - api, demo and source code]

Dear Bob,
Dear Colleagues,

On 2011-04-27, at 17:37 , Bob Lund wrote:

>>> [...]
>> Indeed, this is my opinion as well, that is not giving up on a higher
>> level API, but starting from a "simpler" one and experiment on top of it
>> with JS libraries.
>> Then consider an higher level API, if this is felt as needed, at a
>> second stage.
> 
> I also think this is the best approach. It is not at all obvious how a higher level API that spans the various home networking protocols would look. I think it will take implementation experience to zero in on what, if any, higher level API is best. This will only be possible if we have the simpler API first.

I think I would tend to disagree with your last point that we should only start looking at the service level API after the device level one is fixed.

From my experience I would recommend that we start investigating from both ends in parallel. Start with what application developers care about, and with what the devices can provide. Then start filling the void in the middle. In the projects I have been in which followed your approach, we always ended up with the device semantics shining through at the application level API. This is generally not supporting application developers to the extent possible. Typical requests you get from the app guys: "What am I supposed to pass in as the BufferLeakMode? All I need here is reading the stuff from the URI!". I guess you get the point... ;)

Re. getting the "right cut" for the app developer API, you are completely right that this is induced by experience. As an inspiration, I can offer an API description that is fueled by 10+ years of learning the hard way: http://www.mhp.org/specs/GEM_1_3_JavaDocs.zip (at the risk of repeating myself).


Thanks a lot and cheers,

  --alex