Phil believes that I did drink too much Kool-Aid.

But after reading his post, it seems to me that we agree on most of the points, or maybe it is the Kool-Aid again :)

He says:

The answer, my friend, is ‘yes’. Here’s Gianpaolo’s justification for still wanting to buy an Office licence and run Microsoft’s productivity software on his PC at home:

“… as far as user interaction is concern[ed], I am a big believer in bringing it as close to the user as possible. Why would I ever want the cloud between me and my work?”

I totally agree with this point of view. But instead of going to all the trouble of buying a licence, installing the software locally and then having to maintain it, why not let the cloud bring the application to your desktop whenever you need it? That’s why Adobe developed AIR, why Google has Google Gears, and why Microsoft has developed Silverlight (in addition to investing in Softricity’s desktop virtualization software).

AIR, Gears, Silverlight etc. are not the applications you want to use; they are the containers of the applications you want to use. Once installed on the client machine, they offer a well known 'client side' execution environment that service providers can take advantage of if/when they want to offer a local, often richer UX and/or offline support. Difficulty or easiness of deployment aside, this is conceptually not different than requiring a specific version of Office, .net framework or Firefox to be installed.

Which brings me back to what I was saying in my previous post:

the problem is not running client software (there is plenty of CPU, Memory and Disk at the edge), the problem is easy deployment. I would rather focus time on fixing the root cause (deployment) than working on the symptoms.

This is why A LOT of effort are put into making deployment of these technologies on the client machines easy, fast and not requiring elevated privileges and/or technologies such as Softricity (as Phil astutely indicated as relevant) can play a big role in this model.

I reiterate my premise. Rich 'client-side-software-enabled' user experiences combined with hassle-free-infrastructure back-end hosted services is a far superior model than a cloud-only 'pure' SaaS model. The price to pay for this is deployment of the client piece, but thankfully deployment technologies and/or client side containers are increasingly facilitating all this.

Now if you don't want to call this Software + Services, it is fine by me, but it is what S+S is about.

For completeness, I need to say that rich client UX + hosted services is not the only scenario that the S+S model enables. There are plenty of others, specially in the enterprise where data ownership, regulatory compliance etc. are big decision factors, where S+S make sense but let's stick to the Office / Office Live Workspace scenario that initiated this discussion.

I was about to stop my entry here, but I decided to add another quick proof point that when we talk about AIR, Gears, Silverlight we are talking about software problems not services problems:

If you think about it, AIR, Gears, Silverlight etc.:

(a) they have to be installed on the PC before being of any use, this can be done through 'click here' type of downloads, through distribution deals with OEMs, or 'included' with other popular applications such Acrobat, toolbars etc.

(b) when a new version of these technologies is released, it needs to be redeployed on the client machines, at least for taking advantage of the new features available in the new client side container (e.g. better streaming, access to peripherals...); again either using techniques mentioned in (a) or using some sort of software updater as part of the installed software.  

(c) any breaking changes with code built against older versions have to be handled

(d) penetration of AIR, Gears, Silverlight... on client machine is a key metric for deciding whether to rely on it being present for the client piece or not

(e) ...

... in other words not that different from any other 'client software'.

And again, why would anyone want to go through all these hassles when a 'pure/cloud only' solution would remove all that: because it is MUCH better for users to interact with software locally than on the remote side of the cloud.

I too would love to live in a world where everything happens magically, but until the laws of physics will dictate how things happen on earth :) hybrid models, such as S+S are the best way to mitigate the different constraints.

At the end of the day, I think Phil and I are disagreeing more on the semantic of the term S+S than on the model itself.