Skip to content

Danny Cohen Passed Away

August 29, 2019

Not a theorist but …

[ Internet Hall of Fame]

Danny Cohen was a pioneer who advanced many areas of computer science. He made contributions to computer graphics, networking, and flight simulation, and much more.

Today I want to remember him and his work.

Danny is hard to label. Leonard Kleinrock, said that Cohen “had the uncanny ability to employ his deep mathematical intellect and insight to real world challenges, with enormous impact.” Indeed. This is quoted in last week’s New York Times obit. The Los Angeles Times obit noted his 1981 paper addressed to warring “Big Endian” and “Little Endian” camps.

Obituaries for Danny do not, in my opinion, explain why Danny was so remarkable. For example, he impacted theory even though he never worked in it. Let me explain in a moment. But first {\dots}


I worked with Danny. I was on a government committee that he chaired, years ago. He was an excellent leader, a tough chair, and ran a tight ship. He had been a fighter pilot, perhaps that is why he was such a great leader. On that committee—I will not identify it—he once fired a new committee member as he left the first meeting. Really as he walked out the door, while the rest of us were still meeting. He immediately said that he would not be staying on the committee. Tough. Yes. But Danny knew how to run a meeting.

Danny had a clever wit. One of his favorite jokes was based on a form he had printed out:

The_____________________________technology is a promising technology and always will be.

He then would fill in the blank with whatever you thought was a great technology and hand it to you. Today perhaps we might fill in “quantum computing”—or is that wrong to say?

Danny loved irony. His son revealed that Danny had joined the Flat Earth Society. Danny was initially rejected—no scientists allowed. After he got in—this time he did not list his profession—he framed the certificate of membership, the rejection letter and the Flat Earth Society’s map of the world, complete with an emblazoned “Australia Not Down Under..”


A key contribution that seems to be missed in obits I’ve seen is that he helped create the VLSI revolution. MOSIS is the Metal Oxide Semiconductor Implementation Service. Danny created it, styling it after the earlier work of the pioneer Lynn Conway. Ten points if you know what VLSI stands for.

MOSIS allowed researchers to actually fabricate their designs: to convert digital plans into chips that worked. You designed the chips yourself, encoded in a strange digital file format. Then you sent the files to MOSIS. They then placed you design along with other designs onto one wafer. A wafer held many designs. The wafer files were finally sent to a commercial foundry. There they were converted from digital into silicon. When MOSIS got them back they broke the wafers up and sent you the chips that you designed.

This was magical. The beauty was we could make working chips without having to have our own foundry. At that time frame many companies in the computer business would boast that they had they own captive foundry. Few could afford the huge cost of having their own fabrication foundry. Danny liked to say:

A captive foundry was one that captured you.

His point was that if you had your own foundry, you probably would be forced to use it, even if it was not the right type for your needs. Captive indeed.

Since 1981, when it started, countless projects have been fabricated. MOSIS was a brilliant idea that made it possible for many to build working VLSI chips. We did that more that once while I was at Princeton University. While at Princeton my students and I sent several designs to MOSIS. For example, Dan Lopresti sent in a design that we discussed before here.

The ability to make chips was a terrific motivator. I believe that the excitement of being able to make your own devices helped move the field forward. Without MOSIS, without Danny, the field would have advanced much slower.

Open Problems

Perhaps the lesson of MOSIS is not lost on us. Today there are “MOSIS” like systems that allow even theorists to execute physically real quantum computations. This is indirectly thanks to Danny. We owe him much.

2 Comments leave one →
  1. August 29, 2019 11:04 am

    ⭐ cheers for a NYT obit on a computer scientist, a not a common occurrence. also note that the outsourced foundry idea is still very big even with multibilliondollar companies, eg Apple does it probably with multiple suppliers, although the exact details are probably a trade secret. also google has been making its own chips like tensorflow wrt AI advances, and intel is now going in that direction also (after years of Nvidia pioneering it). chips + their fabrication are a huge strategic advantage and china has major initiatives also. there are rumors of insecure and/ or counterfeit chips being thwarted by the US defense department. it would be cool to see a modern survey/ history on this subject, its very dynamic/ complex. worked on an EE prj in early 1990s that was similar, it was a bit-serial multiplier that the local college was planning on fabricating via mentor graphics designs (iirc), dont know if they ever managed to get it fabricated. my only grad level class, got an A, but a time consuming + stressful experience, esp putting in more hrs than the other grad students as an undergrad 🙂 😮

    • rjlipton permalink*
      August 30, 2019 11:47 am

      Dear vznvzn:

      Thanks for the comment. Did not know the extent that foundry issues still important. I do think the effort in making things real is important. There is something about making stuff beep and ding…that is work…that is cool.

      Thanks again


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s