5,694,593 is another example of the patent office granting a software patent for an algorithm that had been in use in various forms, including the one explicitly called out in the patent, by a lot of organizations for a very long time.

Its some prof who is curious to see how google actually does their query distribution - cos its going to be one clever bit of magic that does it.

Is this a new trick? Using vaguely related patents to probe a

competitors confidential business methods?

Brin and Page's paper "The Anatomy of a Large-Scale Hypertextual Web Search Engine" was published in 1998, so a year after this patent was filed. This is the paper that talks about the notion of 'barrels', in which the index is effectively segmented. The idea being you can invert each index independently so you don't need a single machine with the memory/processor power to invert the whole index at once.

The side-effect of this is also that you can distribute your query across multiple servers too. So sounds like there could be a basis for the patent infringement. But 'backrub' as google used to be (backrub.standford.edu) was probabaly around before this patent.

-Matt

Oh, please. This is USPO we are talking about. They aproved the patent for "Method of swinging on a swing", which described someone swinging on a ordinary two rope swing sideways by pulling the ropes towards him one after another.

USPO only re-examined that patent, after it had become a public joke among patent solicitors. You could propably get a patent for hot water in USPO.

There should be some legal clause stipulating that you cannot wait to file a patent lawsuit until the offending company makes lots of money on said patent. That is what this guy is doing. I wish the judge would just say "Screw you".

How can anyone know that an idea hasn't been used before? It is extremely unlikely for the patenter to actually be the first one to use an algorithm. Many go undocumented due to trade secrets or the true developers don't care to spend significant resources on a faulty patent system. It is impossible to verify that a patent is novel, period.

It is especially ridiculous for person X to claim ownership of an algorithm when persons A, B, and C actually come up with the same idea independently without knowledge of X, possibly even before X.

Then the government grants X exclusive monopoly rights to an algorithm, the problem is that this adds very little value to the community at very great cost. Even if the patent turns out to be invalid, it can cause a great deal of harm to developers.

Why grant software algorithm patents when someone else is perfectly willing to implement a custom solution at a lower cost without using any of X's work. The basic idea behind patents is to spread ideas and knowledge. This is noble, however anyone who's read a patent knows how cryptic and useless they are to actual developers. It's a failed cause and everyone knows it except those who profit by them by draining "intellectual property" out of the developer's domain. When the patent expires, nobody's going to gain any value whatsoever from these now public domain obsolete legal documents.

These days we have far better resources than patents for learning about software technologies: books, technical magazines, "the intenet". In fact I don't know any developers who ever consider patents as a source of information (this is the "value" they were intended to provide.)

These artificial monopolies on software algorithms just never made any sense, but today with the viable alternatives available it never made more sense to ditch software patents than now.

Terradata, for one, was using this approach in the mid '80s. Their original hardware formed a tree structure of task assignment nodes with database fragments as the leaves. A query was decomposed by the task tree with each sub query being sent to the leaf containing the database fragment it would process.

Large ordered tables were split over all the leaves, so sub queries could be run in parallel. The task tree assembed the final result set from sub-query results. It was blindingly fast on most queries or short transactions but lay down and died if you asked it to do a join over two large tables.

This patent sounds like its been invalidated by a case of extremely prior art. Terradata was shipping boxes over a decade before either Google or Jarg were incorporated. It would be surprising if their engineers didn't know how a Terradata system worked.

I've been thinking about similar distributed databases for some time. There is nothing new here that someone with a knowledge of database and high availability computing couldn't come up with in a couple of hours.

A patent is supposed to describe something novel, not something that is so obvious that nobody has yet bothered to patent it.

"Using vaguely related patents to probe a competitors confidential business methods?" sounds like agreat idea - could I get a business process patent on it?

of why software patents should not exist,

It's not a new trick. $DAY_JOB is currently being by our biggest competitor (using patents, of course) in an attempt to get us to disgorge our sales and support methods.

"There should be some legal clause stipulating that you cannot wait to file a patent lawsuit until the offending company makes lots of money on said patent."

Er, there is. Look up "promissory estoppel" sometime. Basically, if you know that somebody is doing something that they shouldn't, but you let them get away with it anyway, you estop yourself from doing anything about it in future.

Google should file to dismiss the case on the basis that (1) the patent is invalid, on grounds of obviety and (2) the plaintiff is being a vexatious litigant.

I have to agree fully.

I do very small scale development from time to time. Could be a batch file to automate parts of a system maintenance, or some small program for a specific job where I can't find something suitable on the net.

For what might be 10 mins of coding on a freebie one off, I could spend hundreds of hours making sure that every routine is NOT patented somewhere. And yet, if I don't make sure that my code is not going to infringe someone's patent, I run a very real risk of ending up in some fairly serious trouble..

Fortunately, the stuff I write is so obscure that it probably won't matter if some routine for getting today's date from the computer is already patented - no one other than me is likely to know the program exists, unless the customers employ someone else to manage their weird systems.