Was it more of his idea, and he handed it off to you? How did that relationship work? MIKE BURROWS: What happened was that he never bothered writing it up. I learned about it from him as a graduate student.
And the years ticked by after that. And he never wrote it up. At the time when I was a student, I thought that it was just one of those things that grad students learn.
And then eventually, I realized, no, this was really special, and the world ought to know about it. When he did it, he did not think of it as a practical algorithm. He thought of it as an algorithm to use to calibrate other algorithms. It wasn't practical for him, because he didn't have a really fast implementation.
And computers at the time were significantly slower. And so it didn't really go fast enough. So I decided that what I should do is write it up myself. And in order to do that, I had to do something. And so I worked with him to make it go fast Casinoslots New Zealand.
And we came up with ways to implement it efficiently. COLT MCANLIS: So I've got to ask you, in the 20 years since this algorithm's invention, what has been the most amazing thing you've seen it applied for? I mean, it's been everywhere, from the Linux operating system to a protocol on the internet.
What's impressed you so much about its use so far? MIKE BURROWS: Well, I suppose the most unexpected thing has been its use in DNA sequencing. It puts together the fragments of DNA that have been sequenced independently very efficiently into a combined whole. But there have been many other advances around it, particularly in ways to do the sorting efficiently, also variance of the transform, one that sorts only limited-length contexts. It turns out that Michael Schindler of the Technical University of Vienna discovered that if you sort the context up to any finite length k, it's still invertible as a transform.
Unfortunately, the inversion takes twice as long, so it's not popular. But it still works. David Scott and Yossi [? Azah ?] managed to find a bijective variant of the transform, where the size of the transform string is exactly the same as the size of the input string. COLT MCANLIS: So you don't need the row index at the end?
MIKE BURROWS: Yes, that's correct. It's a bit more involved and a bit slower to do the compression, but it still works. And that was another surprise for me.
COLT MCANLIS: Give me a little bit of background here. Where did you actually publish the paper originally for this? MIKE BURROWS: There's a funny story about that.
We first of all sent the paper to the Annual Data Compression Conference. But they rejected it. And there were no comments about why they rejected it.
So I wrote to them and asked why they had rejected it. And they told me that it was their policy not to explain why they rejected papers. So we just published it as a technical report. The next year, people at the same conference actually asked me to submit the paper again so that they could publish it. And I said, no.
Meet Our Service!
We can write for you any kind of research paper or other assignments!