Get email delivery of the Cadence blog featured here
One challenge with standards is the desire to avoid unknowingly incorporating patents into standards in a way that gives the patent holder a monopoly to go after everyone using the standard and demand unreasonable licensing terms.
When I was at VLSI Technology, I got involved with patents relating to the main 2G digital mobile standard GSM. When the GSM standards were being created, nobody was very aware of patent issues and as a result a number of patented technologies were incorporated into the standard without it really being regarded as important. For example, Philips had a patent on the specific vocoder used. Once GSM became a huge success, these patent holders wanted a royalty on each phone sold. This became a big issue since there was no patent pool, each royalty was separate, and the sum total demanded was quite large. I think that was the first time that the concept of essential and non-essential patents was defined. An essential patent is one that is inherent in correctly implementing the standard. For example, you don't have a choice of vocoders in GSM, the standard specifies it, and so inevitably any compliant GSM implementation violated that Philips patent. There is actually also a grey area, patents that are not technically essential, but where there is only one cost-effective way of implementing the standard even though the standard doesn't explicitly mandate it. But officially these are non-essential patents.
We were worried at VLSI that we were violating lots of patents and had some exposure. But it turns out we didn't need to worry. Almost every patent was actually for a mobile phone doing something and so it was not possible for us to violate the patent directly since we only built chips, not phones. It is always a good idea when patenting something to go as far down the supply chain as possible (so phones, not chips) since it is a lot easier to get $1 out of someone building a $200 mobile phone than someone building a $20 chip, even though the patented artifact might be in the chip (like that vocoder). However, since we built baseband chips and sold them to the second-string players (so not the Motorolas and Nokias who already had their own chips), we needed to advise them on what their risks were using our chips to build phones, what licenses they might need, and so on.
Then, in the late 1990s, Rambus managed to get a number of their patents designed into the DDR (and other) memory interface standards. There is a long history of Rambus joining and leaving JEDEC, the memory standard group, being guilty of fraud, that being reversed on appeal, some manufacturers paying royalties and others fighting them in court, bad faith negotiations, and more. The Wikipedia entry tells you more than you probably want to know. But for many years, the primary source of Rambus revenue was those patent royalties.
The upshot of all of this, especially the Rambus saga, made standards bodies very aware that they risked being gamed. If a company could get their patented technology incorporated into a popular standard then it could be a license to print money. IEEE in particular did a lot of work on its patent policy and the new rules came into force just recently on March 15.
The IEEE rules are intended to ensure that there is transparency about whether essential patents are being incorporated into a standard and, if so, ensuring that the patent can be licensed on reasonable terms (or that everyone knows that it will not be). The big issue they want to avoid is unknowingly incorporating an essential patent into a standard and then, once the standard is in wide use, the patent holder demands expensive royalties. Of course, there is also a desire not to incorporate patents into a standard unless they are available for reasonable licensing and, when there is a choice, the patent licensing situation gets taken into account.
Or in IEEE-speak:
To make sure that everyone is aware of the issue, every IEEE standardization working group meeting asks the participants whether they believe that any of the technology being standardized is subject to essential patent claims, that this must be made known to the entire working group and recorded in the minutes. If so, then the chair of the working group asks the patent holder to fill out a Letter of Assurance (LOA). There is an escalation procedure if either the patent holder does not fill out an LOA, or if the LOA indicates that licensing assurance is not being provided. In those cases, it goes upstairs to the IEEE Patent Committee.
One key sentence in the LOA is:
Note that this assurance applies, at a minimum, from the date of the standard’s approval to the date of the standard’s transfer to inactive status and is irrevocable upon acceptance by the IEEE-SA.
A company cannot promise to license on reasonable terms and then change their mind and switch to unreasonable terms. LOAs are irrevocable. Read the full text of the LOA (which may not be modified apart from filling out the form).
Take a look at this complete slide deck tutorial on the IEEE patent policy.
Previous: Andrew Kahng on PPAC Scaling Below 7nm
Next: FD-SOI: Is It Really a Thing?