It seemed pretty straightforward based on the press release headline: “Synopsys and IEEE-ISTO launch industry body to evolve Interconnect Modeling Standard.”
Always interested in filling yet another in my seemingly endless supply of pockets of ignorance, I made a couple phone calls to understand a bit more about what was being standardized.
And in so doing, I waded into something of a minor hornet’s nest. Actually, that’s a bad metaphor, because, had I done that, I would have been badly stung. As it is, I sit here, more or less no worse for the wear, as the stinging goes on around me.
Is this some huge sleeper issue that’s going to dominate the headlines with salacious, bitter, biting comment and counter-comment? Alas, no, probably not. But it does shed some light on the sometimes-murky world of standards-setting. And, more specifically, on when a standard is a Standard and when it isn’t.
Let’s start with some basic facts. Synopsys has a long-standing parasitic extraction tool called StarRC. It requires as input a description of the process stack so that the parasitic implications of the circuit topologies can be computed. These process characteristics are stored in a file having a particular format. In Synopsys’ case, this is the ITF format. Because lots of people have used StarRC over time, lots of people use the ITF format.
Of course, ITF isn’t the only format out there. Each tool-maker has a format, and, as if that weren’t enough, foundry TSMC also has its own format called iRCX. So what’s happening is that Synopsys is “opening up” their format under the auspices of the IEEE-ISTO organization. A Technical Advisory Board (the IMTAB) has been convened with a number of players in the industry.
Now, depending on whom you talk to, either
– TSMC is happy about the ITF format being opened up or,
– TSMC learned about ITF before doing iRCX and did iRCX anyway, or
– TSMC did iRCX only after requesting that Synopsys open up ITF, and, when Synopsys declined to do so, only then did TSMC do its own format, only to be further annoyed now that Synopsys has opened up ITF after TSMC did all that work.
Which of these is true is open to conjecture at this point since the one viewpoint not included here is TSMC’s, since they declined to comment. Rumor and innuendo will have to suffice.
So, other than maybe some ruffled feathers, on the surface it pretty much looks like any other standardization effort. (And it’s fair to say that even a normal standardization effort may involve more than a few ruffled feathers.)
Until you talk to some (OK, at least one) of the major players not joining the IMTAB – Cadence in this case. Their Group Director of Product Management, David Desharnais, takes no issue with the ITF format itself: “It’s a good format.” Of course there are several good formats, but ITF could do just fine. And they don’t have any disagreement with the fact that there is demand for a standard for this file format.
Where things become contentious is the means by which the standard is being set; Cadence wants to be part of an open process where each member gets a vote, and the IMTAB process doesn’t work that way. Which seems a bit strange, since such one-member-one-vote processes tend to be a hallmark of standards efforts.
And that’s where a closer look at IEEE-ISTO helps. The IEEE imprimatur can be a golden seal of approval for electronics standards. They’re not the only standards body – there are several others – but everyone has heard of IEEE, and if a standard was done through IEEE, then it’s more or less considered solid. Some standards groups will even set a standard and then turn their own standard over for IEEE ratification to get that extra ring of credibility.
So the IEEE brand is pretty valuable when it comes to standards. And it’s easy to assume that the characteristics with which IEEE standards are associated would accrue to any standards effort bearing the IEEE name. Like IEEE-ISTO. But, as it turns out, ISTO is a completely different organization with a different charter.
Drilling in a bit further, there are two key characteristics of organizations that set standards. One is the fact of legal protection against charges of collusion. Ordinarily, it would be illegal in the US and other countries for competing companies to sit down and agree to do something a certain way. (Especially if that “thing” is setting prices.) So it is illegal for all the major players in an industry to decide to hold a meeting one day and collaborate.
Therefore the first value of a standards organization is to provide a legal framework within which it is legal for competing companies to agree on things.
The second thing they do is provide a process by which standards are set such that no one can really railroad the process or exclude input or participation from any entity desirous of joining the effort. This gives a level playing field, a “one-member-one-vote” structure that gives standards their credibility.
The first element, the legal structure, will be necessary for any effort of any kind that involves the collaboration of competitors. Without such protection, it’s not collaboration; it’s collusion. The second element, that of the standard-setting process, is not legally required. It is typically imposed by, for instance, ANSI. If a standards body like IEEE has ANSI accreditation, then it has passed a high level of scrutiny, further bolstering its credibility.
Getting to specifics, then, the IEEE Standards Association (IEEE-SA) provides both the legal structure and ANSI accreditation to the standards it approves. IEEE-ISTO provides the legal structure, but not ANSI accreditation.
What ISTO does is provide a service to organizations that come to them. As described by their Marketing and Business Development Executive, Peter Lefkin, ISTO doesn’t have or enforce a fixed process; they review the proposals brought to them by organizations. Each such organization will have a process for getting their work done, and ISTO will review that with them.
If the proposed process appears abusive or if it lacks robustness – let’s say the process has the smell of a thinly-veiled attempt to maintain proprietary control but to add a veneer of “openness” while, in reality, shutting competitors out – then ISTO may decline to facilitate the activity. They have declined more than one such proposal.
In the specific case of ITF, Synopsys has put in place a process where they effectively retain ownership of the ITF format but have convened an advisory board to guide the future development of that standard in an open manner so that the format can be used by multiple tools makers and foundries.
So why would they do this instead of going the normal IEEE route, particularly since TSMC appears to be opening up their proprietary formats through Si2 (and presumably iRCX would follow the current efforts on DRC and LVS formats)?
Well, let’s take a realistic look at how full-on standards bodies can work. Let’s say you’ve got a well-established proprietary format or protocol and you turn it over to a standards body. Given that donation, you no longer own the standard. You own your own format, but the standards body can do what it wants with the proposed standard, including change it.
Why would they do that? Well, depending on the situation, the politics, or your viewpoint, it could be because:
– there was a weakness in the original proposal and they improved it;
– standards bodies hate to be rubber stamps and have to change something to justify the time they spend in tropical conference rooms doing the standards work; or
– a competitor just wants to muck things up; if the standard doesn’t match any legacy format, then all players are on an equal footing, and the guy that donated the format loses the benefit of the installed base, and, if you can successfully kill backwards compatibility, then you’ve got a standard that no one wants and no one ends up using, and, as a competitor to the donor, you have now neutralized their efforts and can get a good pat on the back from your management.
So to avoid that eventuality, as an alternative to simply turning your golden technology over to the whims of a standards body, you try to keep control of it somehow.
Like, going to ISTO instead. Synopsys Marketing Director Robert Hoogenstryd doesn’t articulate things in quite such a cynical fashion. But he avers that, with something as widespread as the ITF format, you really want to avoid disruptions. At present, such disruptions might be to competitors’ benefit. But if ITF is taken up by more than just Synopsys, then at some point no one benefits from disruptions, and, at that time, maybe it’s possible to send the standard to IEEE-SA for a proper anointing. Not that he’s promising that… But it could happen.
In addition, Mr. Hoogenstryd points to the typical standardization time for IEEE-SA (long, by his estimation) and the backlog at Si2 (a couple years, by his estimation), and a more flexible way of doing things through ISTO starts to look pretty attractive. For the record, Mr. Lefkin states that “… [standardization] speed isn’t necessarily a differentiating factor” between IEEE-SA and IEEE-ISTO. ISTO efforts can be slow, and SA efforts can be quick.
What it boils down to is that IEEE-SA is a body for setting standards. IEEE-ISTO is really more of a way of managing ecosystems, possibly involving “de facto” standards. (A “de facto” standard sometimes means a non-standard that everyone happens to use, and other times it means a non-standard that someone wants to position as a standard even though it isn’t.) If a standard goes into ISTO as de facto, it comes out as de facto. Its content may have changed, but its status hasn’t.
So what does this mean for ITF? Well, Synopsys is obviously going forward with the IMTAB. Other players will decide to join the IMTAB or not. Issues raised by TAB members do not have to be addressed by Synopsys, so, in theory, Synopsys could simply ignore everyone. Mr. Hoogenstryd reminds us, however, that the Liberty format has been managed in this way for a number of years now and says that “Synopsys has been a good steward” of that format. So it’s not unreasonable to expect that they’ll be equally good stewards of ITF.
Meanwhile, Si2’s president and CEO Steve Schultz notes that the window is now opening for contributions to the OpenDFM Coalition’s effort to include parasitic extraction as the next step in their efforts. He sees it as realistic that they could have that part complete in less than a year. But what’s not as well understood is the fact that they’re dealing with a level of abstraction above the specific file format.
In fact, Schultz stresses that “IMTAB isn’t a conflict or problem; it’s a feeder” into the Si2 process – or at least he would like it to be. He’s hoping that Synopsys will contribute ITF as one of a variety of formats that could form the basis for the Si2 work, along with TSMC’s iRCX, Cadence’s ICT and any other relevant formats. The result of the Si2 work could then be complementary to ITF.
So it all boils down to two efforts here involving parasitic extraction; one at Si2, with the OpenDFM Coalition, and one at Synopsys, with the IMTAB. There’s likely to be some positioning over the next year as the two groups continue their respective work. Whether the results turn out to be truly complementary remains to be seen. You can bet, however, that there are a number of customers and others that will be secretly praying that we don’t end up with two competing standards – that is to say, a non-standard.