In your example, that is O(1). A computer can add two numbers in constant time. How could that be two steps? It is obviously only one operation.
In general, you can count steps as whatever you like. For instance, you could say it takes O(n) CPU operations or O(n^2) file seeks. But then people use the word "time" as a proxy for those. You'll hear "constant time" or "order-n time".
If patent workers are overworked, why don't we replace them with a certificate of origination. You file it, it takes two weeks, it's dated, and it describes the work. Then, if you need to sue someone, you then have to pay to prove that they are infringing. Reverse the burden of proof and delay the cost of approving patents.
Second, my biggest problem with software patents is that you can't build the thing from the description. If you look at a patent for a device, there are schematics and descriptions of how it works. Software patents just say "the user can click a button and it automatically buys a book". Where are the detailed drawings of how the web server connects to the payment system which connects to the distribution center? How do you scale that up to millions of users? That's where the magic is.
Patents are supposed to give the inventor a reasonable period of time to profit from his work. After that time period is over, it becomes public knowledge, so everyone can do it for free. Software patents do not give away the knowledge that it actually takes to build the system. Instead, they typically just describe the tiny bit that is apparent to the user.
The problem that I see with the certificate of origination idea is that then the patent trolls will file for literally millions of certificates of origination. Then they'll just send you threatening letters asking for a settlement and you have to do the costly vetting process to decide whether or not their certificate of origination has merit. Most likely you'll just settle.
A variant on this contributes to the problem. The mechanism is that person 'a' 'invents' X while working on something doesn't think its all that special, person 'b' also 'invents' feature X but they work at a place that has incentivized filing patents so they file a patent on it.
Now some time later person 'c' also invents X and gets sued by person 'b's company. (or the holding company that bought out person 'b's company's IP portfolio).
Engineers have been told to literally "file a patent on everything you do, regardless of how novel or patentable you think it is, let the lawyers figure that out." further, to protect themselves they pay 'bonuses' for filing. Its a strategy and it pays dividends in the current climate.
So I understand Marco's and others frustration here, I would be interested in clearly bogus patents that have been filed post 2005 (my thesis is that starting in 1995 the PTO was hit by a zillion new patents and it wasn't until 2005 that the examiners had become trained enough to recognize the obvious ones and discard them) I've done some expert witness work and found that the prosecution histories of patents show a significant improvement in the examiner's responses as you move from 1995 to 2005.
If there are many defendants, they could work together to find prior art. Crowdsourcing might also be an option - for example, ArticleOnePartners.com pays cash rewards to individuals who find prior art.
Mandatory arbitration (prior to trial) might also be an interesting alternative, especially with the burden of proof on the plaintiff.
The lawyers for the patent trolls are smarter than that, they aren't going to start with dozens of defendants, they are going to start with one or two, bully them into settling, and then use those settlements as precedent when they go after 4 other people... and hey, their patent must be valid, because look at these two guys who are licensing it already.... then they go after the next group (expanding the net each time like they did in those old 'she'll tell two people' shampoo ads).
Precedent counts for a LOT when it comes to resolving any legal issues, even if the precedent is settling, and the trolls know that and use it.
Settlements might make it easier to bully others later, but they do not provide legal precedent. Only an actual court ruling does that. And as senior/superior a court as possible.
My intention was more "why haven't desktop apps copied the best parts of GMail and made it irrelevant?" But here are the features I'd like to see in a desktop client:
Conversation view (one line per conversation in folder view, all emails in conversation shown when viewing the conversation, and the emails are sorted by most recent message, etc)
Fast (fast search, fast email viewing, no pauses while waiting for emails to download, etc)
Simple interface (single screen which changes views)
The thing I'd like most that gmail doesn't have (even with offline gmail) is to have my mail stored locally and be able to use it while I'm not connected.
I would add "conversation view" to the list. It's much better than threaded view, even though it's only slightly different.
I have lost emails in other clients due to the fact that I was viewing them in threaded view and the conversation was really old. They just didn't show up in the first couple of pages of emails and I missed them.
But the question wasn't meant to be about which features make GMail better. It was supposed to ask why a desktop app hasn't copied the features.
Yes, it is amazing how it changed the way I viewed email. But why hasn't a copycat mail client come about, with all of the advantages of running native, come about to surpass it?
There is another problem with software patents that is usually not talked about: they don't present how to build it. Most patents (outside of software patents) present enough information for a competent person to build the device (or whatever) themselves. The idea is that your idea is protected for the duration of the patent, then when it's over everyone else gets it for free. Software patents usually don't do this. I cannot implement, for example, Amazon's "Tabbed browsing" from the description in the patent. I still have to do the hard work myself.
I believe that we should issue something much simpler than patents: file a very brief description of your invention which is stamped by a trusted agency. It is only evaluated in the case of conflict. That is, if you think I am infringing, you bring it to court, present your originality paper (which dates your idea) and then you have to prove that it is not obvious and that it is useful.
This procedure puts the burden on the claimant, not on some random guy who happens to figure out an idea. It also does not require much work until there is an actual conflict.
http://www.faqs.org/docs/artu/ch09s01.html
More data (and data structures) and less code. It's very common in Lisps and other homo-iconic languages.