People will use native apps particularly when web apps just cant provide the functionality needed, but web apps will grow into the large majority, slowly improving and eating out the market from the bottom.
its taken the web a very long time to take out flash's market share so it could take a very long time on mobile as well, however the web is progressing rapidly these days so it may not be the long drawn out process we have seen with flash.
Personally, I'm looking into doing Qt development next year.
Mobile is special, in that the answer to the question of "would less processor/memory/storage overhead be better?" is probably going to be "yes" for the foreseeable future. It certainly will for the next year.
This gives you a very different tradeoff between developer productivity and code efficiency than you would have on a desktop machine, much less in a server farm.
At some point batteries will be so good, and processors so efficient, and memory so cheap that the overhead of not running natively will go from "it doesn't matter except for some games and stuff" to "it doesn't matter". And compilers and development environments will be able to amp up developer productivity without sacrificing power efficiency. So daleharvey might get the last laugh. Just not next year.
That being said I already spend the majority of my time using mobile optimised web sites / web apps and they havent had a noticeable effect on my battery vs native apps, so while I am making it a "prediction", its really already happened, I just think the trend will continue.
I only target iOS as mobile devices, so my focus maybe be slightly different from others (android is catching on a lot around here with my friends, and nobody is picking up blackberry).
For graphical big projects, I tend to write in Scheme and compile down to iphone. It's nice to have the flexibility and comfortable toolset.
I do a lot of small quick-and-dirty projects as mobile javascript apps, and they tend to grow over time, but never too big... until recently, I've been thinking about wrapping these in some sort of native-layer so that I can possibly distribute them via the app-store (honestly, most of the native stuff I used to slug through objective-c to get is available with the new iOS Safari stuff).