Salesforce places bets on HTML5

Yesterday evening my time, I had a conversation with Parker Harris, co-founder Salesforce.com. I always enjoy spending time with him in large part because I always come away with the sense I've learned something genuinely valuable. This was no different. Here are the highlights:

Salesforce.com is betting on HTML5

Salesforce.com has not exactly been a mobile leader. It likes iPhone and iPad but that was pretty much it. It's a problem for all enterprise vendors. In an earlier email conversation with Loic LeMeur, CEO Seesmic, he was bemoaning the fact he has to develop for different flavors of Android. I remember when you had to develop for umpteen flavors of UNIX. The mobile market is even more difficult because it is moving so fast. While developing for Nokia and RIM would have been 'must do' four years ago I can't think of any large dev shop that actively entertains both or either. (SAP still develops for BlackBerry but for how much longer?)

Seesmic itself recently announced it had dropped BlackBerry client development for Twitter, much to the relief of its hard pressed dev team. I know from my own experience that trying to get video onto BlackBerry is almost a no go - apart from YouTube. Nobody cares about BlackBerry. But until I'd spoken with Harris, I was wondering where Salesforce.com would place its bets. Now we know. It's HTML5 all the way. Whether that's a good thing remains to be seen. On balance I sense that Salesforce.com has this right. It just makes sense when you're a cloud player.

While talking about mobile Harris noted that he sees a shift away from the Apple AppStore for business, pointing to the new FT.com application as an example: "The FT doesn't want to pay Apple every time someone wants to get its subscription app." Thinking broader, it's another reason why I think SAP should develop its own appstore.

Identity

Continuing CEO Marc Benioff's recent new mantra about social business, Harris is devoting attention to identity. He talked about the fact when Salesforce.com was conceived, the infrastructure was built such that your business was ringfenced from all other tenants in the multi-tenant architecture. He argues that the cloud is more trusted these days and that trust allows Salesforce.com to open up. He describes this as a combination of chatter plus your business social identity that tracks things in which you are interested. In essence it will allow connections between people across tenants. This is another difficult engineering task because Salesforce.com cannot allow those connections to break the security it has established. This sounds a lot like what TIBCO has been doing recently but again, we'll have to wait and see what the company shows at the upcoming Dreamforce event.

No planned downtime

The last time we met Harris said that he was hopeful the company would have no planned downtime at some point. Those plans are now a lot clearer. Salesforce.com pushed some 500 releases last year. All but three of those are what you would call patches and were achieved without customers losing access. The three main annual releases required that all users be taken off line for a few hours while Salesforce.com spun up the new version. When you're operating at Salesforce's scale that is a problem. Harris says the solution is in sight but does not want to make too big a deal of it until they've destruction tested it - expected in the October timeframe. I don't blame him because this is a major engineering task with important development implications and many potential pitfalls. If the company achieves its goal then it will be an important differentiator. One to watch.

In-memory database

I had become a tad confused about Salesforce.com's position on in-memory database (IMDB.) This is one of SAP's hot buttons at the moment and there's a lot of expectation that SAPS's HANA will deliver serious value. The other year, Harris was talking about IMDB but that conversation evaporated and I thought that IMDB was dead. It's back on the table. This time, Salesforce.com is building its own (code named Vampire) rather than acquire. Harris says the company needs to hold on to its SQL (Oracle) based system for transactions but that having a third party IMDB would have introduced too many complications. It is early days and it may be another year before we see much that's tangible but the company is hoping this will for example allow it to provide high speed complex analytics on stock portfolios. The way Salesforce.com is approaching the problem means they won't be using a columnar database. I'm not convinced. The columnar database approach is proving very interesting, even in the OLTP space but I can understand why Salesforce.com prefers to keep with its row based Oracle system. One thing that's not in the development roadmap is map reduce. Colleagues see this as a vitally important part of making IMDB usable over the longer term. Harris did not dismiss it altogether so again, watch this space.

Push upgrades for Force.com

Many observers have wondered why Force.com has not been the slam dunk developer platform success that Salesforce.com had hoped. Part of the reason is that the way it is architected, Force.com cannot provide push upgrades for applications. This means that developers are forced to upgrade all of their customers manually. Hardly desirable in a cloud era and a serious cost drain. Salesforce is addressing this and expects that to be available in the winter time frame.

Oracle

No conversation with Salesforce executives is complete without some mention of Oracle. Since we're now inching closer to a Fusion release it is apropos to hear what the company thinks: "Fusion doesn't work does it?" said Harris. I'm not going to argue with that!!