Although I agree we should keep a GWT compiled version in mind, I
would suspect that this effort should prioritize making the best
server and api's.  Many custom clients should be able to communicate
with it.

Personally I want a perl client.  But I'm crazy that way.

-Brian

On Dec 5, 8:38 am, Michael K <[email protected]> wrote:
> > The thing to be careful about is how many classes you expect to load into a
> > browser based gwt app. Unless your target audience is running bleeding edge
> > chrome, having a lot of classes is going to lead to sluggish user
> > experiences through gc abuse.
>
> This is an interesting point. I must admit I'm not very familiar with
> GWT yet (though I'd like to change that). I'll keep this in mind.
>
> I don't think there would be a hell of a lot of classes, and they
> probably won't be very complex classes. But there could be a lot of
> instances if you have a large wave open. I suppose that an object pool
> might help here, but there would still be a large memory footprint in
> that case, which could still slow things down..
>
> On the other hand, since GWT is being used for the Google Wave web
> client, I'd expect it to be optimized quite heavily (if not yet, then
> eventually), including the memory management.
>
> > Are you taking the wave robot api as a touch point for your api design
> > ruminations?
>
> I did consider it, yes. But the robot api is very heavily based on
> events, which I'm not sure is the best option here. For one, the
> primary input of a robot are changes to the wave made by other
> participants. But the primary input for a client are (usually) the
> actions of the user. Further, there is no robot api implementation on
> FedOne (although there is a basic agent API, which is somewhat
> similar), so this may require a lot of extra work. And it could
> actually lead to more classes (if not more objects) since you would
> need both the document model and the event model.
>
> I do agree that the event model as used in the robots api could be
> useful for a client. And it would certainly be interesting to be able
> to run robot code on top of a client backend. Actually, the client
> already uses a listener interface for wavelet operations, so this
> could be extended..
>
> I'll have to give it another look when I get to that point.

--

You received this message because you are subscribed to the Google Groups "Wave 
Protocol" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/wave-protocol?hl=en.


Reply via email to