WEBVTT 00:00.000 --> 00:12.960 Hi everyone, so next talk, we have a geometries spin-ellist here who will speak to us about 00:12.960 --> 00:20.080 how they found the V4 tape and upload the data to the UNIX history repo. 00:21.040 --> 00:31.920 Hello, good morning. As you know, UNIX has defined the way modern computing works. 00:32.720 --> 00:38.960 It was developed in the World Famous AT&T Bell Laboratories that is a place associated with 00:38.960 --> 00:46.000 no less than eight Nobel Prizes and three touring awards. And yet it started the life out of 00:46.000 --> 00:53.360 a failure. AT&T, General Electrics and MIT will be building a system called Maltics in the 1960s 00:53.360 --> 00:59.920 that was a bit too ambitious for the capabilities of the time. So AT&T pulled out of the project, 00:59.920 --> 01:05.920 it wasn't going to well. And can Thompson, other scientists found themselves without a subject of 01:05.920 --> 01:12.640 studies. So can Thompson later joined by Doug McKillroy, Jefferson and Dennis Richie went on 01:12.640 --> 01:19.600 and developed the first non-named system on a PDP 7. That's some very interesting ideas in it. 01:19.600 --> 01:24.880 And based on the success of that system, they created the funding proposal to Bell Laboratories, 01:24.880 --> 01:31.760 they needed to do that even in that well-funded place for a more powerful computer. Initially they 01:31.760 --> 01:39.680 asked for a PDP 10, which cost about half a million dollars at a time. Bell Ups told that they cost, 01:39.680 --> 01:46.000 and at the end they got the PDP 11, which cost a tenth of that price. And maybe the lesser 01:46.000 --> 01:52.880 capabilities of that system influenced the direction of Unix. Over the 1970s, a large group of 01:52.880 --> 01:58.160 people, you see them here at the later age, they were very young and powerful. At the time, 01:58.720 --> 02:06.000 they developed on and created a series of Unix editions. These were numbered following the 02:06.000 --> 02:12.480 editions of the manual that came out each time. In various states, you see here the seventh 02:12.480 --> 02:19.040 so-called Bell Ups research edition from the first one to the seventh one. So source code was 02:19.040 --> 02:25.840 not versioned even at the last edition there was no version control system, but the manuals were 02:25.840 --> 02:32.400 versioned through their printing and adding an edition word, word on it. Not a lot has survived 02:32.400 --> 02:37.600 from the first edition, so for the first unnamed one, it has survived in printed form and people 02:37.600 --> 02:43.040 have also already already made it run. In others, when you see you have some parts of commands, 02:43.040 --> 02:48.560 some parts of the kernel. At the fourth edition, we had just the documentation. 02:50.000 --> 02:56.720 Which brings me to 1974, and this letter from Kent Thompson to Martin Newell at the University 02:56.720 --> 03:01.120 of Utah, which says that we are delaying the delivery because we are on a lot of copies 03:01.200 --> 03:07.760 of the documentation, but there will soon send you a system. Newell, you might know him, 03:07.760 --> 03:12.880 he is the person every one computer scientist, specializing computer graphics. He is the creator of 03:12.880 --> 03:20.560 the famous Utah team or the benchmark used for 3D rendering. So it seems at some point there was a 03:21.840 --> 03:29.760 version that could be shipped from Bell Labs to Utah, and something like this status of 03:29.840 --> 03:36.400 the American, it's not the original tape, but I want to show you to you for those who are not familiar 03:36.400 --> 03:46.880 with the computer tapes, got shipped from Bell Labs to Utah. Fast forward to 1828 of July in 2025, 03:46.880 --> 03:53.360 and the storage closet of Robert Rich's flat research group at the Maryland Engineering building 03:53.440 --> 04:00.160 you will see here, where Alex and American researchers associate in the group found the tape 04:00.160 --> 04:06.560 among the documents of Jayle Pro. The findings was widely reported, even on broadcast TV. 04:07.760 --> 04:15.680 To, so was there anything on the tape? It was important to read it, but also it was one tape 04:15.680 --> 04:21.680 we never had access to it before, so it was considered very valuable. So to avoid high altitude 04:21.680 --> 04:28.000 cosmic radiation on flights and airport scanner damage, lab members jodering, you see at the 04:28.000 --> 04:34.560 center in Thalia Archibald, you see on the left and there took 11 hours drive to the computer 04:34.560 --> 04:43.920 history museum in California. There it was read by Archivist Alcosso of bit savers, famous archiving 04:43.920 --> 04:50.320 site, you see him here at work. To minimize tape where he used the system developed by computer 04:50.320 --> 04:57.760 historian Len Shashik, so he first read the tape not as a digital bit, but as a 1.6 gigabyte 04:57.760 --> 05:05.520 of analog data using a multi-track analog to digital converter. This means that a lot of hardware 05:05.520 --> 05:12.080 that you see here on the left that happens inside that is implemented normally inside tape readers 05:12.080 --> 05:18.080 can be implemented in software and thereby the tape can be virtually read again and again with various 05:18.320 --> 05:26.960 settings so that can be restored. You see here the waveforms from various tape tracks and 05:26.960 --> 05:32.000 here zoomed in to see how they can be seen and various elements how they can be adjusted. 05:32.640 --> 05:40.960 This was successful and now if you go to archive.org using Shostek's read tape program, 05:40.960 --> 05:47.520 this 1.6 gigabyte analog data were converted on 2.5 megabytes of digital 05:47.680 --> 05:54.480 the format which is compatible with a image emulator and both are now available on archive.org. 05:54.480 --> 06:00.720 Think of it, tape is 2.5 megabytes about the same as a single picture we take with our mobile phones. 06:01.840 --> 06:08.480 What's the legal status of this tape thankfully in 2002 called data which on the IP rights 06:08.480 --> 06:16.800 to Unix at the time wrote this letter which if you squint well enough looks like an open source 06:16.800 --> 06:23.840 license. So effectively it can be uploaded on negative and shared and built upon and run and so on. 06:23.840 --> 06:29.200 And then people started using it for example Brian Rodriguez developed this erotic 06:29.200 --> 06:35.200 ebook the Unix 4th edition source code commentary where he analyzes the entire kernel and how 06:35.200 --> 06:41.680 it is built, how you can run it on an emulator following a very influential book from the 1976 06:41.680 --> 06:48.080 by John Limes from the University of New South Wales which taught many generations of computer 06:48.080 --> 06:55.200 science students the design of a real operating system kernel. Also Angela Papenhoff who 06:55.200 --> 07:01.680 gave a very interesting talk at 39c3 if you weeks ago made it run on a pdp 11 and you later 07:01.760 --> 07:09.520 and also published a tar file with the tapes contents. I employed the file to add the tapes 07:09.520 --> 07:16.000 contents to the Unix history repository. What is that? It's a repository on GitHub on the URL 07:16.000 --> 07:22.560 you see there that contains a history of Unix from its inception so the first unnamed version 07:22.560 --> 07:28.640 until today for a lot part of it it traces the 730 search editions work that happened at the 07:28.720 --> 07:36.400 Berkeley the BSD editions 36 BSD and ends up with the 3 BSD modern editions from the very complex 07:36.400 --> 07:43.360 and long history of Unix it traces one line they want you see here on the in orange. It contains 07:43.360 --> 07:50.640 snapshots of the research editions and values Berkeley editions then commits moved from 07:50.640 --> 07:56.880 CCS this was the source code control system used for many years at the Berkeley and then finally 07:56.960 --> 08:07.040 normal free BSD commits as they move from CGS to subversion and now git. The way it is constructed 08:07.040 --> 08:16.320 allows us to run git blame on it here I do it on times on dot c and amazingly we see on the 08:16.320 --> 08:24.000 modern free BSD times on dot c lines that were apparently written by Dennis H.1979 actually some 08:24.000 --> 08:29.200 of you may realize that those lines shouldn't be there because what was released by Berkeley 08:29.200 --> 08:35.920 was supposed to be a complete clean source open source implementation of Unix so it's a mystery to 08:35.920 --> 08:46.480 me to to add the contents of the tape to the repository what I do is I what I did was first 08:46.480 --> 08:52.160 of all associate files with actual authors based in documentation and the information that was available 08:52.160 --> 08:58.800 at the time even oral histories and things that people who were there told me now and I got 08:58.800 --> 09:04.160 the timestamps from the actual timestamps of the tape and this each timestamp is a separate 09:04.160 --> 09:11.360 commit. Let me demo you how the system works you can go to a Unix before dot dev and see 09:11.360 --> 09:16.720 the running this not work at it but I find it pretty cute cool and it shows a Unix system that's 09:17.280 --> 09:23.760 and what we expect so if I type on the command line I can change the directory to TMP 09:23.760 --> 09:29.680 I run the line editor it's what happens when you press on VIM a Q I append lines and I write 09:29.680 --> 09:33.920 the small main function with a hello world that you would expect. 09:37.200 --> 09:44.560 I press dot 2 finish my input I write the file to hello dot c and then I can compile it 09:45.440 --> 09:51.840 if I list the files in the directory c and a dot dot and if I run it press to get hello world 09:51.840 --> 10:03.040 from code that was written at that time. So this is really important because now we have a complete 10:03.040 --> 10:07.760 it's the first complete Unix distribution we have the kernel the commands we already had the 10:07.760 --> 10:14.560 mantpages and the c compiler. This version produced many interesting things structured programming 10:14.560 --> 10:19.760 so kernel was implemented in a language called the new B at the time you can guess which language 10:19.760 --> 10:28.160 this was 7800 lines of new B and just 1000 lines of VDP 11 assembly. Provided I think for the 10:28.160 --> 10:33.360 first time a language independent API so you see here the pipe system called documented both in the 10:33.360 --> 10:40.160 way you call it from assembly language R0 and R1 RDP 11 registers and from c to get an array 10:40.160 --> 10:46.160 of two file descriptors. It used the data structure definition and reused through header files 10:46.880 --> 10:52.080 it provided the device driver abstraction through special files many of the files are 10:52.080 --> 10:59.760 completely nonsensical today speech synthesizer and the spider interface but the driver interface 10:59.760 --> 11:06.480 the way this abstraction was implemented is still available today and previous editions but also 11:06.480 --> 11:11.360 similar design on Unix and even on Windows if you try to write a Windows device driver. 11:12.000 --> 11:17.520 Also it had a properly typeset manual you see the fourth the third edition is typewriter 11:17.520 --> 11:24.400 the other one is a typeset. It had a snowball and implementation a dynamically type string and 11:24.400 --> 11:32.160 pattern processing language that has influenced the or compared and thereby Python 1600 lines of 11:32.160 --> 11:38.080 C and tried to find out what the way it was written it was just a quick entertainment for 11:38.080 --> 11:44.640 again and the application that survived was a program that solves the instant insanity page and 11:44.640 --> 11:51.040 here you see it written in snowball. What are the tapes contents regarding dates what I did that 11:51.120 --> 11:58.080 created the rank of diagrams about from the files of all editions that have survived and if you 11:58.080 --> 12:04.800 and as you can see the fourth edition the files match many of them match the manual page some 12:04.800 --> 12:10.320 are associated with the later date but they seem to be relatively close to the manual date. 12:11.440 --> 12:18.160 I also wrote a script using a git to git blame to see what code was inherited from previous 12:18.320 --> 12:26.080 editions and as you can see the fourth edition in had some parts of the third edition and also 12:26.080 --> 12:32.800 the fifth edition inherited a large part of the fourth edition. I also looked at the evolution of language 12:32.800 --> 12:38.880 used because it was a time C was being used more and more and you can indeed see here that the 12:38.880 --> 12:45.280 fourth edition content is a significant part of C code something that continues and gets expanded 12:45.360 --> 12:50.800 until the end of the decade and I think we can unfortunately be fortunate to say that it 12:50.800 --> 12:53.840 survives until today. Thank you very much. 12:53.840 --> 13:16.960 So we have a couple of minutes for some questions. Thank you very much. I volunteer for 13:16.960 --> 13:22.080 computing history group in Newcastle UK and about 10 years ago the retiring head of school 13:22.080 --> 13:28.480 handed me a tape and he said this has got unix on it I think. I wouldn't get too excited because 13:28.480 --> 13:33.440 I'm fairly sure it's it's well known it's V7 or something and obviously the next thing I need to 13:33.440 --> 13:37.760 do is go back to him and ask him because he's still that I can still kind of contact him but 13:38.320 --> 13:43.760 assuming there is something of historical interest on it what would you advise our next step to be? 13:44.640 --> 13:50.640 Contact the computer history museum to get further advice we may be witnessing history here. 13:51.600 --> 13:54.400 Thank you. Thank you very much for bringing this up. 13:58.480 --> 14:05.360 Riddle or tape is tricky they can get destroyed so please don't touch it. For example the tape 14:05.360 --> 14:10.320 right if you doesn't have any plastic parts, cups, tants and so on in order to minimize 14:10.320 --> 14:29.520 it weren't there. When converting the binary data when converting the 14:29.840 --> 14:35.360 data to binary how did you make sure that there were no errors in the transformation? 14:36.160 --> 14:42.560 So when I didn't convert the tape I just uploaded it on GitHub but when the binary data were 14:42.560 --> 14:49.920 converted how to ensure there were no errors there as understand the CRC checks in the tape 14:49.920 --> 14:55.520 reading mechanism and then people actually looked at the tape contents to see that it was okay so 14:55.600 --> 15:01.920 2.5 megabytes is fine and indeed we found some people find some blocks that were mislade 15:01.920 --> 15:06.080 and manually patched them to collect everything. Amazing, cool. 15:11.760 --> 15:17.280 We are on time but let's take one more question and then we move to the next talk. 15:26.320 --> 15:34.880 Hi, thank you for the amazing research. The Bose Extenders came out in 1998. I was wondering 15:34.880 --> 15:40.880 to what extent does this very early incarnation comply to the later Bose Extenders? What is 15:42.480 --> 15:48.080 to what extent does this feel like a real modern units? To what extent does it feel like a 15:48.080 --> 15:55.600 real modern units? There are differences, for example I typed CHD, rather than CD. Many of the tools 15:55.600 --> 16:02.080 are already there, command options are missing and some tools are not there yet. 16:02.080 --> 16:06.480 And the kernel like the system pulls that were later the fines and the Bose Extenders, 16:06.480 --> 16:10.960 they're still being developed at this point again. Thank you for that. So what happened with 16:10.960 --> 16:18.480 the interface and the system calls? There is this paper you can find that documents and 16:18.480 --> 16:23.760 the repository on GitHub that has a time line of all unique facilities, command system calls, 16:23.760 --> 16:29.760 when each one appeared and when some of them also move the way. And there you can find the exact 16:29.760 --> 16:34.960 timeline and see what was there and how it compares to the 7th edition and later ones. 16:34.960 --> 16:38.160 Very impressive. I'm going to look at that up. Thank you. Thank you. 16:40.960 --> 16:44.160 Thank you again.