At last, whatever the mobile widget engine chosen, it seems that for now, the effort is not worth it compared to the really small reachable audience.

Still, it would be nice to have a comparison between those 3 ones, are there others?

I could understand better if you objected to java widget platforms generally

I believe *that* would be hasty. I’m sure there are good arguments against widget platforms (and widgets themselves) written in Java and I would definitely agree with some of them, but I don’t believe that WidSets is representative of the potential of even a ‘typical’ Java widget platform, nor that Java is necessarily unable to provide a great user experience given a competent developer (team).

I just didn’t understand what you found to be “heavy”?

I thought I did specify that: Responsiveness to user input, apparent network performance (viz. widget loading times) and of course start up time — on all three, at least on my phone, Yahoo! Go 3.0 beta is faster. Perhaps some optimisation is at play there which makes things faster for Yahoo! Go on my phone. Regarding the network performance, it seems to me that WidSets is overtaxing the CPU of the device by trying to load and run a large number of widgets at once. In any case, my intention was not to provide a thorough diatribe with all that’s wrong with the client; it was just a simple expression of the impression I got that more or less conveys my disappointment.

Finally, regarding your ending comment where you write “WidSets development is a god-send”. While I can accept the conclusion, given that there are few (if any) widget platforms for most phones out there and WidSets definitely has its niche from a developer point of view, I don’t accept the premise though — which is that I was hasty in my criticism; specifically, I don’t believe it has much to do with what I’m criticising here: despite its problems, WidSets development would be excellent for anyone trying to develop small, self-contained ‘widgets’. Sadly, this does not make WidSets neither a useful platform for users (for the reasons mentioned in the article), nor a technically sound piece of software at least on my phone — and this is perhaps the only part where I erred above: that is, I generalised based on the performance of a single, high-end Nokia phone. Still, given that WidSets is developed by Nokia and that — again — on the N95 8GB the apparent performance of Yahoo! Go 3 β is significantly better than that of WidSets is probably a good enough reason to assume that lower-end devices will provide a worse user experience.

I was referring to Yahoo Go 3. I had also previously tried Yahoo Go! 2.0 and, as you say, 3.0 Beta is an improvement but still slower than WidSets for me.

Of course loading times is not representative of quality and I would make no such claims (in mentioning loading times I was reponding directly to your comment that the application is “really slow to start”).

Being able to write a slow, midlet in 4k is obviously a given. I just didn’t understand what you found to be “heavy”? I could understand better if you objected to java widget platforms generally (as Russell Beattie does) but I’m just surprised at the polarity of your opinion in this case – Yahoo Go “excellent”, WidSets seemingly terrible? Surely not.

Anyway, you obviously know your stuff, but I think your judgement may have been a bit hasty in this case – I come from a J2ME background and WidSets development is a god-send.

WidSets seems to repond slightly faster to user-input than Yahoo Go.

Are you referring to Yahoo Go! 3.0 Beta? Yahoo Go! 2.0 was considerably worse, in my opinion, so it’d be useful if you qualified your statement with a version number. If you did try Yahoo! Go 3.0 beta, then I can only express my surprise. I found it much faster than Widsets in terms of responsiveness at least on my phone.

In terms of loading times, a fresh install of Yahoo Go took about 15 seconds to load up and WidSets took about 17 seconds – not much difference there.

Loading time is not representative of an application’s quality or responsiveness. Many applications cache data upon loading and this may add to a perceived delay. Specifically, I’m quite certain that Yahoo Go! loads quite a lot of data and code upon startup, whereas the Widsets client seems to load Widsets after the application has fully loaded — in which case you cannot really compare the two; furthermore, Widsets loads widgets over the network, which further complicates any attempt at comparing the two.

In terms of WidSets being too heavy – WidSets Java client is around 200k whereas Yahoo Go is about 750k.

Again, being ‘heavy’ has little to do with application size. You cannot compare the size of the ‘jar’ files and have a verdict. I can write a midlet that’s only 4K when packaged, but is much more demanding (be it due to bad coding style, bad use of resources, algorithmic complexity or combinations thereof) than one that’s 1.5MBs and full of pretty images and other resources.

I guess if you’re from a web dev background

I’m not, actually, although I’ve worked in that field too over the yeras. My background is firmly rooted in software engineering.

But for me it seems that with a Java-esque script language comes a lot greater potential for interesting widgets than those based solely on web standards.

Agreed. However, given the constraints of current mobile devices — even the high end ones — the need for responsiveness, usability and resource control, I believe that something akin to Javascript coupled with — perhaps — an appropriately optimised, lightweight object model providing access to network and system resources beyond XHR could be beneficial. That’s a wholly different discussion though and one we could have in another post. :)

Thanks for your comment Will.

I guess if you’re from a web dev background, there’s more of a hurdle to developing for WidSets. But for me it seems that with a Java-esque script language comes a lot greater potential for interesting widgets than those based solely on web standards.

However, I agree with many of the sentiments in your article – any widget interface should be near-instant loading if possible. Also, an overhaul of S60 is definitely needed, and a widget-based dashboard would be very cool indeed.