Variation is essential for competition and competition is essential for evolution.
At first glance, you could argue that we never need any form of standards. However, as things evolve towards more uniform, standard, commodity like components then they enable the evolution of higher systems which in turn evolve.
Evolution begets genesis begets evolution.
Evolution begets genesis begets evolution.
Hence we have a natural conflict within technological and economic systems, namely between variation which enables evolution of a system versus standardisation of that system which enables the evolution of higher order systems which consume it. In other words;
Variation is essential for evolution of a system but lack of variation is essential for evolution of higher order systems.
Standard nuts and bolts does limit variation of the nut and bolt but it enables rapid development of higher order machines. Standard electrical interfaces does limit variation of electricity supply but it enables rapid development of higher order systems that consume electricity.
The history of technological progress involves a constant resolution of this conflict.
But the question is how do you choose a standard and when do you know it's right to choose a standard? The answer is generally that you don't, fortunately we have a mechanism which can help us to determine the answer - it's called us or more aptly the market.
Our ability to make a reasonable choice depends upon the size of the group involved (i.e. a large market trumps a committee) and experience (i.e. examination of actual use trumps a survey of people who aren't involved). Markets because of number of actors (both consumers and suppliers) involved in active use including those creating higher order systems are pretty good means of identifying the standards. What the market chooses is commonly called the de facto standard.
The only time any form of interference is needed is when it becomes clear that the market is failing i.e. some form of level playing field is not being created for a maturing act. At this point, Government's often have to step in and ideally this should be through open standards in order to create a more functioning market.
Hence, in the case of Cloud then we're almost certainly too early for such interference. The market seems to be functioning well, de facto are emerging and there seems to be a high level of competition. All looks good and healthy. In the case of document formats then a level playing field for a mature act didn't seem to be forming and hence the introduction of a open standard like ODF seems a necessity to correct this.
Making a choice on standards is not easy but must be carefully balanced with the user needs, the maturity of the market, how well it is functioning and the state of evolution of an act. Interference should be minimised especially when it is clear that there exists high levels of competition and a probable de facto (e.g. with Cloud IaaS then the AWS APIs have multiple open source implementations from Eucalyptus to CloudStack to Open Nebula to various distributions of Open Stack). Choice should be kept to a preference unless it is clear and the Government can make an informed decision.
Alas standards can be gamed to the advantage of a firm. There are many historical precedents for companies using standards processes to undermine others. I tend to view the use of committees to create standards as a last resort best avoided. Do we need standards in the cloud? Yes but that's already starting to happen in the market, so let it happen. There are however places Governments can help such as simplifying and unifying legislation such as data protections rules. Which brings me to my next topic ...
On Patents
The reason why I mention this whole topic is due to a conversation on Patents. There is variation in the patent systems of different countries and that's actually good for competition. We are far from knowing what a good enough patent system is and attempts to force a uniform patent system are just as likely to create problems as resolve others. The patent system itself suffers from what I consider to be a one size fits all mentality in that the length of terms are generally fixed. You can never find a single length of time which fits all markets i.e. what is too long for the software world is too short for the pharmaceutical.
A better option, which I've argued for over the last decade, is to allow the length of term to be variable upto an upper limit (say 25 years). For every patent, the length of term should be set to "slightly more than the likely time for independent discovery of this invention". Hence if you had patents in the software world, then they should probably be of the order of a few weeks. Whereas discovery of a teleportation system certainly merits consideration for the whole hog.
But how do you determine the right mechanism for this? How do you determine the right length of time? You don't. You let the market fight it out and only interfere again when what is created is not suitable. But in order to do this, you need to set the conditions carefully and to create a conflict which resolves the issue.
In the same way the conflict between evolution today (variation in underlying component) and evolution tomorrow (variation in higher order systems based upon a lack of variation in underlying component) helps create de facto standards for the market through competition then you need to set up similar conditions for patents.
Hence your patent process, a trade between society and innovators, could start with the award of a patent with a term of "notional length of time equal to [time awarded] and initially determined to be a maximum time for independent discovery with any actual time to be determined by market conditions". Hence I get my patent for five years say but I know that the amount of time awarded might actually vary below this amount.
Now, we need to set up the conflict as the company (and its lawyers) will have an interest in gaining the maximum time and so we have to a mechanism to encourage others to minimise it. The best way to do this is to put the whole value generated (which increases over time) up for grabs.
An example would be: "Should the length of patent time exceed the actual time for independent discovery, then any party can claim against the patent holder all direct and indirect revenue along with any costs incurred associated with the invention. An equal sum to that awarded to the party must be paid to the state."
This creates a hard choice for patent holders, the longer you hold onto the patent then the larger the risk becomes that someone will sue you for exceeding the time of independent discovery and what is at stake increases. You can't set up a subsidiary to sue yourself as any successful win means an equal payment to the state.
For a patent troll, along with the normal battle of trying to apply your patents then as soon as you've won then you potentially open yourself to counter suits depending upon the strength of your patent. The cost of which could be the original claim which is then paid to an third party and the same again to the state. Hence, winning with trivial patents could become very costly. And yes, you would enable an entire market of lawyers acting as Anti-Patent trolls and suing any and everyone for holding patents beyond what they can demonstrate is a likely time of independent discovery but then that's the point, to create a natural conflict.
The conditions set up as above would encourage patent holders to use patents (as is the current situation) but also to retire them early and to apply them only for a reasonable length of time. Now, I'm not proposing the above be used but I'm setting it out as an example that you can create conditions which enables the market to determine the answer.
What you want to avoid is pre-supposing you know what is right (i.e. 20 years as a length of term) and instead create a natural conflict which enables the market to decide. It's the same with standards, there's a natural conflict which de facto solves and the only time you should directly interfere is when the market is failing.