Interview 2 – Future Nostalgia Project
Place: Digital Preservation Lab, Cambridge University Library
Interviewer: Leontien Talboom
Interviewee: Keir Fraser
Date: 6 June 2025

Interviewer: There we go. I think maybe we can start by you giving a little bit of context about why you developed the Greaseweazle or how you got into it and all that sort of stuff.

Interviewee: Well, I suppose it's the original interest, so I got into computers when I was a child and I started off as ZX Spectrum, but no floppy disks involved there at the time. 

Interviewer: No floppy disk.

Interviewee: But I got an Amiga sort of around 1990 and played some games, but I was quite interested in what makes it tick and got into Assembly language and directly accessing the hardware. Machines back then were simple enough. You could understand the whole thing top to bottom. It kind of kindled my interest in systems programming, which ended up being my career in various respects. 

So there was this whole stack of manuals for the Amiga computer. The slimmest one was the hardware reference manual, which told you how to programme the hardware. For whatever reason, one of the bits of hardware that most intrigued me was the floppy disk drive. I don't know, I think that's pretty weird. 

Interviewer: That is pretty weird, but we all love a floppy disk here.

Interviewee: Actually, it is quite a bit. I think that super-interested that many people, but I was intrigued by the tiny diagrams and stuff, which were part of the reference manual. I spent a lot of time writing my own disk routines.

Interviewer: Oh, cool.

Interviewee: Disassembling professional software, looking at their disk routines, removing their protection routines, that sort of thing. Anyway, moving onward, so the Amiga era ended at the time. I got a PC, I moved to Cambridge, did my degree, did my PhD, did a start-up company called XenSource, which got acquired by Citrix back around 2007, 2008 sort of time. 

I worked there for a few years, then I had a bit of time off. I think this is true of a number of people from that sort of time who were children and picked up beloved particular retro systems. They're going to go, now I'm a professional computer programmer or whatever. I've got a lot more experience I could bring to bear. What could I do with that hardware now and you've got people who you can see it in modern demos. It was kind of, or games or any other software, there's kind of renaissance for some of these old systems. It's mostly people who were kids or teenagers at the time and did stuff and now they're like, well, can I do even better with modern tools and modern knowledge? The answer is generally yes. There's some amazing software and things out there now. 

I then got into, well, can I go back and look at the software from the Amiga and kind of come up with some modern tools, not necessarily on the Amiga itself, but for the PC, for Linux, for Windows or whatever, and produce projects which allow us to preserve or delve into the software of yesteryear for the Amiga specifically still at the time it was. So I started it. I had a project imaginatively called Disk Utilities. It's a pretty generic boring name.

Interviewer: At least it does what it says.

Interviewee: It does what it says. Particularly, it is a bit like the KryoFlux has this kind of closed analyser, which they use for generating their IPF image files of copy protected Amiga software. This ended up being basically an open-source equivalent of that. You could supply a raw disk dump from an Amiga disk, which you might've grabbed off of the Amiga or using KryoFlux or using SuperCard Pro, which was just about to appear around that time or just after and you'd stick it in an analyser. You'd tell it this particular game, mostly games. So it would have a database of formats that you would know about and it would go, okay, well that's Amiga DOS, but except for the protection track or it's Rob Northern's protected disk operating system for the Amiga. It would know this from its little text database and it would then go and convert the raw dump to IPF or whatever and it would check that the dump was a good dump because it knows what format it should be looking at and where the checksums are. 

So I developed that and then as part of that, I kind of became aware more broadly of the range of disk formats that are out there. The Amiga has a very flexible disk controller, so it has lots of different custom disk formats. Then I came to realise, I hadn't realised before, that most systems use this kind of IBM soft-sectored format, right, which goes right back to the 1960s actually, and was used by most computer systems, continues to be used by most. Well, not that floppy disks exist anymore, but I'm still getting to the floppy disks. It was by far the most prevalent family of formats. I kind of started getting a broader view of what's out there. 

So that project kind of carries on and, but it's kind of Amiga specific, but I kind of got to thinking, well, KryoFlux and SuperCard, they're the 2 kind of hardware kind of dumping projects, flux dumping projects I knew about. I had one of each of those. I was like well, most people aren't going to get this hardware because it's like £100 to buy the board or whatever. They're both proprietary, basically. They're not public in terms of schematics or firmware or indeed the software that drives them on the host side. So the whole thing is closed. That's pretty unfortunate for a preservation point of view. 

So I also thought what's the cheapest hardware we could come up with that would be able to do the job. That was about 2018, 2019 sort of time. I'd already got experience of writing firmware because just before that there's this floppy disk emulation platform called Gotek, which it's the form factor of a 3.5-inch disk drive. It comes with some fairly limited factory firmware that I think they got a family of different firmwares, but the main one would be it emulates a PC high density disk drive and you give it high density disk images up to a hundred of them like a jukebox. But with different firmware, you can make the hardware a lot more flexible. So you can use them in Amiga. You can use them in music synthesisers. You could even emulate a 5.25-inch drive if you put a little adapter on it or whatever. 

So I made a firmware for that and that was kind of my first foray into writing firmwares. So was doing Flash Floppy firmware for the Gotek. Therefore, I think also, as well as looking to do an open equivalent of the flux dumpers, it also was kind of a good companion for my flash floppy. So I'm doing the disk side, disk drive side emulation. This would be a floppy disk drive controller basically providing a bridge from modern PCs via USB, via this board, go into the drive and give you the lowest level possible access you can get without actually doing hardware modification to the disk drives or other tricks, which are cool, but impractical for most people. The aim was to make it open and as cheap and easy as possible to build so that more people could dump their disks.

Interviewer: Nice.

Interviewee: So the first kind of iterations of that, we're using little microcontroller development boards called things like Blue Pill or whatever, tiny little things. You could get them at the time for about £2 or something like that. 

Interviewer: Wow.

Interviewee: I came up with, it was a bit kind of jury rigged, but you could kind of put little DuPont jumper cables onto the pins of this board and just directly plug them into the pins on the floppy disk drive and it would often just work. 

Interviewer: Oh, that’s cool.

Interviewee: So for 2 quid now you can dump your own disk because the board has a USB connecter built in and well, you go. So it kind of snowballed from there a bit to actually developing open hardware, but custom board designs with a 34-way connecter on them and a power connecter and more nicely done, but still basically the same firmware. And then sort of adding a lot of it then is the host software. The hardware and the firmware is basically sort of done. I hardly touch those at all, but any time I do spend on the project these days is mainly adding new disk formats or host software features, which are requested from people. I've got this system and I give a whole bunch of format definition files for different systems or whatever and that one isn't featured. So they'll give me an example. 

Interviewer: Oh, they would give you example.

Interviewee: Usually you get a KryoFlux format dump or... Not that these come from that particular hardware, but they've become kind of de facto standard flux interchange formats just because they already existed and they sort of do the job. So it's either KryoFlux or SuperCard Pro image files. They dump those usually using the Greaseweazle, share them with me, I look at them and go, there must be this sort of format and write them a definition file.

Interviewer: Nice. And then you would add that to the library?

Interviewee: It's a little bit just flat text files of a little description language I've got, which describes for a given disk, what the format should look like.

Interviewer: Oh, is that also with the interleaving and stuff?

Interviewee: The uploads do specify interleaving things. So you can use that. So you can get them from, for example, a flux dump all the way to a verified sector file, an IMG file. And then because the interleaving stuff is all specified, you can get from that back to a raw dump or write it back to disk.

Interviewer: Very cool.

Interviewee: Because IMG files don't include interleaving. It's more like a logical sector order dump.

Interviewer: When you started this, were you expecting there to be that many formats?

Interviewee: I suppose because I started off on the Amiga and there with proprietary games and stuff, there are loads and loads of different formats, basically. I spent a couple of years writing dozens and dozens and dozens of track for my handlers and we've probably got 99% coverage and it’s like but that last 1% is a really long tail of … I don't really work on that project anymore, but someone else got access to it. And every so often, every few months you'll add another half a dozen format handlers and it'll support another dozen games or something. I don't know, but probably up to 3,000 of, I don't know, 3 and a half, 3,200 Amiga games or something like that. I don't know the exact numbers, but it's like most things are in there. 

If they're not in the text file, it's because simply no one's added it there yet. It may well be that the track analysers exist, but that particular game hasn't been added to the database. Then more rarely, but it still does happen. This is a completely different protection track or it's this flavour of protection track, but it's been tweaked or something like that. Or this game programmer decides to do their own disk routines and just completely made up track format. You can get a whole bunch of those. 

Interviewer: It’s a completely different thing.

Interviewee: They all have a flavour, but yes, exactly. So that means coming from that and there are similar things or they'd know at the time on systems like Apple too, and Commodore 64 and other two, which have jillions of different formats. I guess I was kind of expecting it, but most systems only have one or two different formats. Largely, and probably from a preservation point of view, you're not probably dealing with copy protection, you're dealing with data disks. So the problem there at least is a bit more tractable. There you're probably more dealing with the quality of the media. Are there kind of hard errors with the disk? Is it possible to get a good extraction of the data? If you can get that, that's the hard bit. That's the hard bit.

Interviewer: Because what we have, what I've really seen, and I've been talking to colleagues across the community, is that the Greaseweazle can help us, again similar, with about 95% of the stuff that comes in. So most people will have used a personal computer that was just regular, nothing special. But what we've been seeing that is a lot of specific business systems and stuff that, because they're not that interesting, or there's not that big of an interest around them because they didn't emulate games, they were used for research, or they were used for word processing, they're not as well documented. 

Interviewee: Oh yes, definitely. Right. Yes.

Interviewer: That's the ones where we were a little bit worried about.

Interviewee: Also in some cases, not all of these old systems do use the IBM soft sectors. There are all kinds of weird things out there. If the disks are hard sectored, which some of the 70s stuff is, they're almost always using their own custom thing. That made the hardware simpler and then there's usually some really simple kind of shift register thing there just to clock bits of encoded data out to the disk or bring it back in. The sector format that the programmers then came up with can be pretty much whatever popped into their head. The byte, or you don't know what the bits order in bytes is going to be, or what the sync word is going to be. You don't know anything about it, I guess. You need some documentation.

Interviewer: Exactly. And it's kind of similar to that kind of 1% tail of a lot of it's covered, but there's a small tail that will go on for a long time.

Interviewee: Yes, that's right. New track formats get trickled into the Greaseweazle project. That's kind of the most interesting bit. If you're like, you can give me some disk dumps and typically, if we can dredge up some data sheets, you get to write a new track analyser and that's nerdily quite fun. I could be writing a GUI or something, I suppose, for the Greaseweazle, but in the low-level track analysers is fun. 

Interviewer: It's kind of solving a puzzle, isn't it?

Interviewee: It's exactly like solving a puzzle, you’re exactly right. If you've got the raw disk dumps and you sort of just start with this, a copy of some other track analyser and you just start throwing bits of new code in there and you can just keep running the dumps you've got through and it goes, oh, it's failed here. You just keep tweaking, tweaking, tweaking it until it all comes together and then you write the … You're right, it's that kind of 2 halves to each analyser. There's taking a raw flux dump and pulling out the logical sectors. There's taking the logical sectors and remaking a raw track that you could write back to disk. So you have to write both halves and then you can test it against it. 

Interviewer: You have to do both.

Interviewee: So you have to do both.

Interviewer: Which is very cool, isn't it? So talking about the raw flux, so within our community, we don't currently have any real best practises around what raw flux we should use. And as you already mentioned, there's a few that were developed by more of the proprietary board such as SuperCard Pro and there's like … Do you have any tips for us as a community around what one to use? For example, I tend to use SCP, which is SuperCard Pro, I think it is. 

Interviewee: That’s right.

Interviewer: But one of my colleagues uses HFE, which is ... I can't remember what it stands for.

Interviewee: HxC, Floppy Emulator, probably.

Interviewer: Oh, that's for that one.

Interviewee: I would go for, so the raw flux dumps are closest to what you're reading off of the disk. It's actually giving you the timing between flux transitions, which are being passed to you by the drive hardware. If you go for HFE, then you've actually converted, you've already converted, done one level of conversion. You've basically extracted bit cells, ones and zeros from the flux timings. 

So the format ends up being potentially a lot more compact, but you've already made assumptions about what the bit timing should be. And if there's any, as there can be kind of marginal, if the disk is a bit dirty or whatever, you can end up with kind of marginal flux time. So you end up running a PLL, this kind of clock recovery algorithm over the flux timings to try and … You're basically, you're expecting a number of different, the bit cells fall into different timing windows. So for high density MFM off of a PC disk, it will be, the timing should be, the flux transition should be 2, 3 or 4 microseconds apart. But what happens if you get a transition that's at 3 and a half, it's almost halfway between the 2.

Interviewer: So then you kind of get an unknown sector, something weird.

Interviewee: I don't know if you've ever, you've probably visualised. I see a lot of people use the HxC software because it's got this visualiser in it where you can bring up a track of raw flux data. You can see a noisy disk because a clean disk has these 2 or 3 clean horizontal bands, which are the flux timings. The timing between transitions is kind of on the y-axis. You can see for a PC disk 2, 3, 4 microseconds. But if there's mould on the disk or something's happened to it, they get a bit wobbly. They almost start touching and if those bands start touching, you're like, well, how do you know whether that's a 101 or a 1001? You can end up inserting extra, shifting bits around basically or swapping bits. 

You get another bite at recovering that marginal data if you've kept it as flux timings versus having already converted that into the bit cells, the 1s and the 0s. So I always think that from a preservation point of view, you should do a flux. You should keep a flux dump, and it should be, say, 3 revolutions. Some people say that this is wasteful and uses a lot of space, but then floppy disks aren't that big.

Interviewer: They’re not that huge.

Interviewee: They're not that huge, even a raw flux dump. Between SCP and KryoFlux formats, personally, I prefer the KryoFlux stream format, even though it's a bit more unwieldy because it's got a folder per track. One nice thing that allows you to do, even without tool support, is you can actually splice together an image from, easily do this via the files, via your PC's file system. You could do a dump of tracks 0 to 10 and 10 to 20. You could do it in chunks and then just aggregate them into one complete dump or you can go, actually, those tracks were a bit dodgy, dump those again and copy them over the top of the old. To get yourself an improved dump. You could theoretically do that with SCP, but you obviously can't just do it by copying track files around. 

Interviewer: No. 

Interviewee: You need a tool to do that for you, which doesn't exist. It won't be that hard to write one, I suppose, or something to do with SCP management thing. But what else do I like about the KryoFlux format? There's historically been confusion about how the SCP format should handle single-sided images. It's an arcane thing, but some tools get confused by that. The fact that it defines the number of revolutions per track as an image-wide parameter. So you can have a different number of revolutions dumped for different tracks. That's just sort of a bit of a shame. So in that sense, I think the KryoFlux format is fine.

Interviewer: It's fine. It's the one to go for.

Interviewee: It's probably the one I would go for. I would at least go for one of those two. I wouldn't convert to HFE. That's already cooking the data to a degree. 

Interviewer: And then you’re already making assumptions.

Interviewee: You're already making assumptions and I think that yes, you should be analysing, you should be trying to extract the contents of this. After all, that's the whole point of the exercise is to extract the data. You want to be able to go back to your source. Even if you've no longer got the physical disk, because that will degrade or has to go back to wherever it came from, that record is the flux dump.

Interviewer: Nice. So it's SCP or Kryoflux.

Interviewee: SCP or the KryoFlux. One of the two. I have a personal preference for the KryoFlux but if you use one or the other of those, I think that's a good place to be.

Interviewer: Especially interesting, I knew that the KryoFlux did it in different tracks, that you get a track for every read. 

Interviewee: That’s right.

Interviewer: But it's quite interesting what you're saying as well, is that if you were to then, I don't know, recover certain tracks, you can just basically overwrite it and make it better image-wise.

Interviewee: Again, it's kind of an arcane format. I believe it's literally the data format that comes off of the KryoFlux. So there's some software that controls it and tells it what track to read. But the actual data stream coming in, that's exactly what comes in off of the firmware. I don't think they cook it at all. I think it's just done literally.

Interviewer: Just that.

Interviewee: The software just writes it to a file. Obviously, they've got an analyser built in as well so they can generate other formats. But the KryoFlux stream format is the KryoFlux hardware format.

Interviewer: But it is the most raw, or the most unmodified.

Interviewee: Yes, it's giving you the flux timings.

Interviewer: And that's what you want.

Interviewee: That’s what you want. Exactly. 

Interviewer: I don't know how relevant this is going to be to you, but I assume that you probably have floppy disk drives at home.

Interviewee: I've got a few floppy disk drives. I've got a lot of 3.5-inch floppy disk drives, a lot of PC floppy disk drives, just because, you get them off of, you look on eBay and, or at least I historically did, you go, at least, I don't know, 10 years ago, it was even easier to pick these and you go, here's a job load of 10 or 20 floppy disk drives. So you pick them up, or you can test them, you can see if they all work or not, and get yourself a stack of verified, at that point in time, floppy disk drives.

Interviewer: Do you do anything to your hardware, to your floppy disk drives to ensure that they work correctly? So just to give a little bit of context, within our community, there's a little bit of a worry of like, so a lot of places will only be able to acquire maybe 1, maybe 2 floppy disk drives. So you can't really do a lot of comparison. Is there anything that you do to ensure that your equipment works the way that you're expecting it to work?

Interviewee: I suppose, since I'm more of the kind of 3.5, that would probably be things like 8-inch disk drives, which seem extremely difficult to get hold of. Most people haven't got those things. If you've got a few disk drives, you can at least cross compare, cross test between them. I'll have some, I'll have a handful of original proprietary disks, which haven't been overwritten, which I can therefore assume were originally written by a well calibrated drive. I can use those to test that my floppy disk drives can read those okay. Apart from that and doing some head cleaning and maybe applying a bit of grease, no, I don't worry too much about getting into the drive hardware itself.

It's kind of a bit more esoteric, but there are interesting things that could be done if you do. For example, someone did a project where they went into the test points on a disk drive and got out analogue signals from the drive head. I can't remember who did that now, but that was for, I think they did an example of Repton or something like that on the BBC computer. It was like a bad disk and you got just garbage if you went through the normal interface for reading out the disk.

Interviewer: Oh, interesting.

Interviewee: You just got noise in that section. It was like someone had actually, I think it was actually a dent in the disk surface or something like that. But they were able to, if you go to the most drivers I think have test points where you can get out analogue waveforms from the drive heads. They would be used for calibrating the drive or whatever using a proper test disk. They're able to get a signal from that and actually recover data that you'd think would be unrecoverable. 

Another thing you can do would be to add a microstepper. So rather than having the drive head do complete track steps according to step commands sent via the normal interface, you could directly connect the stepper to a microstepping controller on a new Greaseweazle, so to say. So that's something that your interviewee on Monday was asking about. 

Interviewer: Yes.

Interviewee: A few people have asked about that actually. 

Interviewer: Oh, interesting they have.

Interviewee: Being able to get finer control, radially speaking. 

Interviewer: Of that stepping. 

Interviewee: So if the disk was written, badly calibrated or whatever, maybe you can get a stronger read signal fractionally off where you'd expect it, basically and maybe you could come up with software that could determine which are the best microstep track rates. That would be quite nice.

Interviewer: That would be quite that, niche but nice.

Interviewee: It'd be niche, because it would require you to modify a drive or whatever. But these niche things are quite fun.

Interviewer: They are quite fun. That sounds actually really fun.

Interviewee: So these are the sort of things that interest me more than … Then again, I don't do a lot of preservation myself but I do like to talk to people who do and find out what formats and features they’re interested in.

Interviewer: Whoa that is interesting. Because I think that's something that our community is not always too sure about because, of course, the retro gaming community, which I think you would classify yourself as part of, is not always, what's the diplomatic way of saying this? It's not always approachable for our community in the sense that sometimes we will ask a question and it's seen as not stupid, but at that level where it's like…

Interviewee: I think it might be different points of view and different interests to some degree.  But the good thing is there's a few open-source projects. Well, there are other open-source projects as well. So, the HxC software that people use these days is hosted on GitHub and is open source and Jeff Del Nero is quite an approachable person.

Interviewer: Yes. And I would classify you because I've heard a number of people in my community ask you questions on GitHub. I know also being really approachable.

Interviewee: It's me. There's FluxEngine, David Given.

Interviewer: Also very good.

Interviewee: This is another nice thing is by making cheap, readily available hardware, other projects will add support for your hardware. So, FluxEngine can drive a Greaseweazle directly or you can feed it dumps you made earlier. But it's got its own, I've stolen some of the former analysers from there and added them to Greaseweazle, but it's got its own twist. It's got a GUI, it's got some file system handlers. Some people would probably find it more friendly to use. It's certainly got some features which Greaseweazle host software still doesn't have. So, it's kind of a case of finding a tool chain, if you like, of software things which work for the job at hand. 

Interviewer: Exactly.

Interviewee: If someone says I want to image Spectrum or Amstrad CPC disks or something, then I'll probably say use Sandisk, which is another completely different piece of software by someone who knows a lot about Spectrum and Amstrad CPC custom track formats and has written a bunch of analysers in their software project. So, some of these things almost get at the trickier end of the preservation spectrum. It kind of relies on domain-specific knowledge. It's kind of siloed, therefore, by people who have a deep knowledge about the particular systems concerned. So, I won't personally get too deep into certain, Apple II, for example. I don't know much about Apple II or Mac, right?

Interviewer: It's its own thing, isn’t it?

Interviewee: It’s got its own thing and the standard there would be to use Applesauce. Who's going to compete with that software? Probably no one. It's certainly not something I would spend my time on, particularly because you need a Mac, but the software is readily available, and you can take dumps from a Greaseweazle and put them through the Applesauce software. So, just go do that.

Interviewer: That's the workflow that a lot of people have started doing with Applesauce.

Interviewee: It's great software. A lovely user interface and stuff, unusually for this area. It's one thing Greaseweazles. Greaseweazle does have a third party interface, I shouldn’t be unfair, but it's not as beautifully integrated as the Applesauce.

Interviewer: No. The Applesauce does a very good job of doing that. 

Interviewee: Does a very good job. It does. 

Interviewer: And the images that it can produce, like the visualiser.

Interviewee: They're extra pretty versions of what the HxC software can do.

Interviewer: It's so nice. That’s interesting, that you need a chain of tools for your workflow depending on what you are trying to accomplish.

Interviewee: I think it'll always be like that. For your community, it's building up a corpus of knowledge about what's out there and also keeping abreast of it, because these projects are … New ones appear, and the existing ones are improved, and just keeping on top of that, I think, is quite an effort.

Interviewer: Is there anything that our community could do to help with anything that you develop?

Interviewee: I think keeping contact would be … I think there are a number of people, myself included, who are interested in supporting floppy disk preservation better, and just having people in that field reaching out is good. Perhaps you have events or conferences or workshops where you discuss things, and I could be interested in that sort of thing.

Interviewer: Oh, that is good to know.

Interviewee: I'm sure I wouldn't be the only one either.

Interviewer: No, but that is good to know. I can bombard you with stuff. But if you are actually interested in that, because we have conversations within our community about floppy disk preservation, but it would be so much more, I think, fruitful in the sense that people like you would be part of those conversations to understand, because you've got such a deep knowledge of this stuff.

Interviewee: Right. So, I've got a particular knowledge from my point of view. But the other thing is that, most people who are making software projects want their software to be used. 

Interviewer: Yes.

Interviewee: I hope I have a good focus on wanting to make stuff that is useful and easy to use for a large number of users. But I've also learnt that what I think users want and what users actually want aren't naturally the same thing. As the programmer, you don't really know what users want, and you develop this thing that works a certain way, and then users tell you it makes completely no sense, or that's not actually what we want. And some of the best features that I've implemented in my various different projects have been user requests, where users actually approach me and been willing to spend the time to iterate on that feature. 

Interviewer: Nice.

Interviewee: Like you go, so is this what you mean? They'll be like, well, yes, but change it this way, or I'd do it that way, whatever. Iterating with users to develop new features.

Interviewer: That's really good.

Interviewee: Is almost one of the quickest ways to improve a software project. Just doing it as the programmer, thinking what would users like, doesn't really work. 

Interviewer: Exactly. Like sitting there.

Interviewee: So you’re not doing preservation myself anyway, so how do I know?

Interviewer: What we want.

Interviewee: What preservation experts want.

Interviewer: I think what we want is a tool that is easily accessible and is interchangeable with other pieces of software, and Greaseweazle already ticks a lot of those boxes. So having something that we can get a real flux stream from. So we've got an FC5025, which you've maybe come across. It's in that machine. 

Interviewee: I haven’t.

Interviewer: It’s a floppy controller that was released, and it can only work with 5.25-inch floppy disks, but it cannot do a raw flux stream, and it can only do about 20 formats. They're all very much more of the DOS formats, the early DOS, which is okay, but not really within a UK setting where there's quite a lot of obscure 5.25-inch floppies around. 

Interviewee: That’s right.

Interviewer: So we don't really have that much use out of it. Maybe it's, again, a little bit of an example of developing something, but then not necessarily understanding what users actually want. We're lucky to have it, and it's cool to have different tooling but it's not what we want out of it, if that makes sense. Sadly, I can't take it out because there's a warranty on that machine. I cannot wait until the warranty is gone, and I can tinker around with it.

Interviewee: You can break it open.

Interviewer: And have a look around it.

Interviewee: So what is the preservation community, and I'm asking you questions. What is the preservation community? How do you share information amongst yourselves? Do you have online forums or?

Interviewer: Yes, so most of us are part of certain member organisations. So one of the big ones within our community is the Digital Preservation Coalition, the DPC. So that's one of them. Then there's ones that are more for a little more niche things. So there's the IIPC, which is specifically for web archiving, which is also something that we love. And then there's one more for software, so the OPF. So we're basically loads of us works for institutions that are members in these networks.

Interviewee: So it's like a world I'm over in at the moment. I don't recognise any of those. 

Interviewer: If you want to, I can send you a little overview of all the stuff.

Interviewee: That'd be super interesting.

Interviewer: Cool. And then there's the big conference that happens every year, which is iPRES, and it's the big preservation conference.

Interviewee: That covers preservation more broadly, I assume, with tracks for different things, as I imagine.

Interviewer: Yes. But there's also a lot of quite niche ones around floppy disk and that type of stuff. There's a number of people within our community that are, one who's very much into file formats around Apple computers, which is really cool. I think the step that we're always really interested in as a community as well is being able to preserve the material, but then also be able to access it, because access is, of course, why would you preserve anything if you can't make it accessible?

Interviewee: Absolutely.

Interviewer: So that's a big one in our community that we do. Again, with more of the common formats from floppy disks, it's quite easy to get access to files. It’s easy but…

Interviewee: Yes, it is another one. These old systems, not only do you have to work out what the sector format is, but you work that out, you get sectored on what file system is it?

Interviewer: Exactly.

Interviewee: You don't know what the metadata layout is, then you can't extract files, so you're still stuck. That's frequently less documented or less easy to work out in the sector format.

Interviewer: Yes, so I think that's the next step. So it's from sector format to actually being able to access the files. Then the next step, and that's also something that's really interesting, is you're probably stuck with a bunch of obsolete file formats that need specific software to open.

Interviewee: Exactly, even above that you need specific proprietary software.

Interviewer: It's a level upon level.

Interviewee: You've actually got to work it all the way back up to the top. You probably need an emulator or something, whatever.

Interviewer: Yes.

Interviewee: So that all feeds into it as well.

Interviewer: Yes, so it is very…

Interviewee: It’s a big stack of stuff.

Interviewer: It’s a huge stack of stuff. I think especially for that stuff from the 70s and the 80s, it's getting more and more difficult because more and more of that documentation is being lost. So it's a worry. Because for a lot of institutions it's already difficult to be able to get a raw flux stream because getting these drives is difficult. But then that's only step one out of 10 steps in getting stuff done.

Interviewee: Yes, rebuilding the hardware or building an emulator. Problem is, those kind of deeply technical things, you need someone with the knowledge and the interest too, right? Which is why it ends up even like the disk formats, the file system format, it kind of ends up being people with a retro interest in those specific systems and long-term knowledge of them who end up developing their own individual software projects to address them, which you then have to find out about and stitch together. And if they exist at all, which for a bunch of 70s and 60s systems they don't exist.

Interviewer: It's not around. Especially because the ones that … and it sounds from you as well, normally the people that I talk to who are deeply invested and deeply interested in it, it either comes from having worked on that system in a work setting or having had one of these computers at home. There's a deep, deep interest in it. 

Interviewee: Exactly right.

Interviewer: Where it's like, if it's just a word processor that was part of a business somewhere, it’s kind of boring, isn’t it?

Interviewee: It's a bit of a yawn fest isn’t it? 

Interviewer: It's kind of boring. But that is sadly quite a lot of the data that we see. Just as a community, we get agendas from politicians or printouts from specific meetings. From a historical point of view, it's an interesting era, but it's not the interesting computer stuff.

Interviewee: Is any of this data protected? Do you have to keep it? I presume that you can't just publish it either, let people have at it because it's like sensitive.

Interviewer: And I think that's something interesting in our community as well. I think a lot of us would just love to put our raw flux streams online and just have at it. Just have a go at it, have fun. But I think because they come from certain politicians or they’ve come in with certain donations that have a donor agreements with them. So they're not open until 50 years after I've died, that type of stuff.

Interviewee: Fair.

Interviewer: Which is far too far in the future for a lot of this stuff, which is fine if it's a book and it's in a stable environment. It's fine if it stays there for a hundred years, even 200 years. But if it's a floppy disk.

Interviewee: It's a bit more of a problem. 

Interviewer: It's a bit of a problem. So it could be that legislation and stuff around opening up raw flux streams will become easier, but not at the moment. 

Interviewee: No.

Interviewer: But maybe in the future, maybe in the future. I would love to give raw flux, either just put them on the internet or give them to anyone who wants to have a go at it.

Interviewee: Maybe they can be shared with particular people. NDA or something like that would be a step in that direction.

Interviewer: And I think there's something really interesting. I think that's what I find really fun about floppy disks as well is that we've got a lot of them with no labels on them.

Interviewee: No labels at all.

Interviewer: No labels at all, so you don't even know what system they've come from.

Interviewee: They never have a label or they might have fallen off.

Interviewer: Yes, there's some of them that have glue residue. So I think there was a label on them. But I find that really fascinating because you don't know what system they've come from. 

Interviewee: No.

Interviewer: But then on top of that, you don't actually know what's on there. It could be treasure, or it could be nothing. 

Interviewee: Nothing. It could be literally nothing, right?

Interviewer: Literally nothing. I know that it gives a lot of people a headache within the community but I do find that a really fascinating point about them, that it's like, you can open a book and know what it's about but not with these disks.

Interviewee: I think there's a whole bunch of kind of enthusiasts who are interested in looking at certain types of things. They get a box of tapes or a box of disks and some people actually reading them all and seeing what's on them. I think the same could be true of dumps.

Interviewer: Dumps or floppy disks.

Interviewee: They're looking at raw dumps and trying to work out what's on them can be quite interesting. Therefore, you need access to them.

Interviewer: But I think there is something interesting there, isn't there? Around like you say, having a network or specific people that can help with that type of work.

Interviewee: So at the end of the day, the community is not that big.

Interviewer: No.

Interviewee: The bulk of the internet is not going to help you. 

Interviewer: No.

Interviewee: So maybe there's a way of bringing more people into the fold.

Interviewer: I think that's certainly something to explore because I don't think there is. Personally, I don't think there will be a different way of actually being able to access this material if those types of relationships don't get set up. 

Interviewee: That’s it.

Interviewer: I had a few questions I had written down, but we've kind of covered that, but maybe we can go a little bit further in it. What would you recommend people to do if they come across a really obscure format of a disk? Would it just be to make a raw flux stream or would… and then send it to you.

Interviewee: Always raw flux stream and they send it to me or .. Probably me but there are others. Me, or the FluxEngine guy, Applesauce guy. There are various people who like working out formats from raw flux. If it's a new, esoteric format, that can be interesting. Even better if we know what system it's from, we can go looking for data sheets.

Interviewer: That is fun, isn't it?

Interviewee: And a lot easier if you've got a data sheet. Yes, raw flux.

Interviewer: Where would you find the data sheet? So is that something that would be published somewhere?

Interviewee: Internet archive or something like that.

Interviewer: Would it be published in the manual, that type of documentation?

Interviewee: For these old systems, often, the manuals were like schematics and actually go into what the sector data format is and stuff. The manuals were super technical back in the 70s.

Interviewer: They're very fun to look through.

Interviewee: Right, at least even give you hints, like how many sectors per track that's useful. You get some hints at least.

Interviewer: So with the Greaseweazle, it's really nice that you have the step function and stuff. So it means that you can use a high density floppy drive to read double density disks. Are there any tips or stuff that you would warn people about if they were to be using that? Because I just use the step function and just assume it works. I think I should rephrase this. Is there anything that you would warn people about when they are imaging disks that they should keep an eye out for?

Interviewee: I don't think you should be shy to dump everything that you can. So I wouldn't even use the step function when doing a raw flux dump. I just dump all the tracks. You can always apply the step later. Like in Greaseweazle, you've got the convert function, which is the read function, except that it works from an existing image rather than accessing a disk directly. A FluxEngine I know has a similar concept. So you've doubled the amount of flux that you've dumped, but it doesn't really matter.

Interviewer: So that would be your tip, just dump.

Interviewee: Dump everything. Get the rawest, do the dumbest, rawest dump that you can and work from there.

Interviewer: And then go from there. Is there any indication, because I've only worked with high density drives, we're quite lucky in the sense that we have high, or high density drives, I should say. So if we've got double density disk, we can read them with them.

Interviewee: So that'd be high density 3.5 and 5.25?

Interviewer: Mainly 5.25. If you were to do a basic raw, I'm just hypothetical here, but if you had a double density drive.

Interviewee: Double density drive.

Interviewer: You had a high density disk. You put the high density disk into the double density drive and you did the dumbest, rawest flux stream, would that still show up as if it took all the data?

Interviewee: That's an interesting question. Obviously it can't write them. Can it read them? You know what? I've never tried because I have high density drives. 

Interviewer: Interesting.

Interviewee: I would expect that it probably doesn't, because there's enough drive electronics that it would be tuned. A lot of drives, at least somewhat tuned for the timings that they expect. Yes, I'd be surprised if it worked. I'll just use a high density drive.

Interviewer: Just use the high density.

Interviewee: Just do that. The other thing with 5.25-inch drives is, do remember that you have to tell the drive what the density is, as opposed to 3.5-inch drives where the drive auto detects the density. You've got this pin 2 thing. It can vary from drive to drive. Which way round is high voltage, double density or high density. You have to sort of work that out. I guess once you know it for a particular drive…

Interviewer: Then it's fine. 

Interviewee: As long as you know what density the … It's easy with 3.5-inch disks, 5.25-inch disks. It can be a bit more difficult because there's no kind of notch or anything on them for the density.

Interviewer: I'm pretty sure there's not. No.

Interviewee: No, I think the double density ones sometimes have a little reenforcing ring on the inner bit of the diskette.

Interviewer: But then it's not always.

Interviewee: It’s not always there. It could be missing. You can't guaranty to therefore be able to tell whether it's double or high density.

Interviewer: Because the thing that I kind of worry about within our community is the fact that I'm in a very privileged position that this is basically my job. Of course, it's floppy disks and USB sticks and hard drives. I look at all the obsolete carriers. But within our community, a lot of people will have a bajillion responsibilities, but also have a responsibility around doing the floppy disk imaging basically. 

Interviewee: Right.

Interviewer: I have enough time to faff around with different drives and ensure that I get the best read and also have an understanding of what high density means or double density means or double sided or all that type of stuff. 

Interviewee: You can probably get a reasonable idea if your dump is, broadly speaking, good or not. Whether you've got the right parameters, whether you've got the density right or whatever by doing a raw dump, and then loading it up into HxC, doing the track visualisation at the flux level, and looking to see whether it looks like random noise, or when you've got those horizontal bit bands.

Interviewer: So if it's random noise, that's bad.

Interviewee: If it's random noise it's bad. If it looks really noisy, well then maybe calibration's bad or maybe the disk is dirty or something. If you've got reasonably well-defined horizontal bands, you've probably got a good dump. You can do that just by visualising, you don't even need an analyser there. That's a fairly straightforward visual check that anyone can do.

Interviewer: Because that is something that I do worry about a little bit, because I can get enough time to spend to look at the floppy emulator in detail and stuff but maybe having some, I think writing some guidance around what you're actually looking at when you open a floppy emulator. Like you say, when it's fuzzy around the edges and that type of stuff, because I think that's really helpful as well. So that's that. 

Now we actually also kind of touched upon this as well. Is there any other software that you would recommend? I know you've already mentioned Applesauce and FluxEngine and other stuff. But is there anything else that you would say this works really well with the Greaseweazle, or this is something that I use that I find really helpful or?

Interviewee: Let me think. Other software projects. I think I've mentioned most of them, as I say. Another one that maybe isn't so useful for preservation though, but is a really nice bit of software, is called DiskFlashback by Rob Smith. So a lot of people want to be able to easily access the contents of the disks or the disks that they dump on a Windows PC. They even want it to just look like a drive letter. The data is drive F or something. So Rob Smith has kind of implemented that for a limited range of, so you can do PC disks, Amiga disks, ST, Avatar ST disks, a bunch of sort of, relatively speaking, more modern, you're not going to be able to put a 70s system disk in and have it do anything, but that's a really neat ... It kind of integrates with Windows Explorer, so that’s on the taskbar, so that it can mount disks directly using a Greaseweazle, or if you've got a disk dump it can mount that, let you modify it, access the files through a pure drive letter, that sort of thing. That's kind of neat. 

Interviewer: That is neat. 

Interviewee: But beyond that, I kind of got these. If I was going to do Commodore 64, then there was a project called, I think, DDR64 that someone in the community wrote. It just happens to be good at generating G60. It's another esoteric C64 specific kind of raw data file, but that's good. It’s probably you just need this knowledge of these kind of lots of different domain specific tools.

Interviewer: It seems there's very system specific tools that are really of interest. Is there anything yourself that you would love to develop for the Greaseweazle?

Interviewee: I would like to make it more user-friendly and add a GUI onto it. I don't know if I'd be any good at writing GUIs. I've never done that before. But what always interests me is new formats and new hardware hacks. So stuff like, say, microstepping. Doing a prototype new board and hooking that up and seeing how microstepping works would be fun. So kind of low-level stuff like that. But equally, if I'm always interested in adding new features, which are actually going to be useful to people who are actually using it.

Interviewer: I will say to everyone in my community, Keir really likes hearing about stuff that we need. Because that is something as well, where does the name Greaseweazle come from? Is it like a play on Catweasel?

Interviewee: Yeah.

Interviewer: It is? Okay.

Interviewee: So Catweasel was this Amiga specific disk dumping hardware. It's basically a play on that. A greased weasel. So it just sounds good. Everyone remembers the name. So it turns out to be silly but...

Interviewer: It's silly but it's good, isn't it? 

Interviewee: Silly but good.

Interviewer: I love that it’s purple. 

Interviewee: I’m not generally good at names though.

Interviewer: Well, I wouldn’t say that. It is a good one.

Interviewee: There's no disk utilities that one.

Interviewer: No. That's a lot more catchy, isn't it?

Interviewee: Yeah.

Interviewer: I just love that the board, when we ordered our first one and it arrived and it's a purple board with a weasel on it. It's difficult to explain to anyone within the library who's just like, "Wait, what's that?"

Interviewee: It’s a greased weasel.

Interviewer: That’s great. Because that was something that the interviewee on Monday mentioned as well, the Catweasel. So he said that he sometimes uses the Catweasel because it can detect a little bit more information about the disk before you image it. I don't know exactly. I don't have a Catweasel.

Interviewee: No, to be honest, I've never used it either. One of those interesting bits of hardware that I was always interested in. That is true that there are other projects which will try to automatically determine, sort of automatically determine the disk format and then do a sensible dump. Whereas the Greaseweazle attitude, at least at the moment is more of the like, you really should know what. You need to know. This isn’t going to tell me what system it is, what type of disk it is and then I'll just do the right thing. It doesn't do any guessing for you. 

There’s this kind of IBM scan format I added where it will go and scrape off basically all of the IBM soft sectors that it could find on every track. Some people like using that. It probably does work all right because it will give you warnings if there are checksum errors and things like that, sector data errors. You could perhaps theoretically find a disk that was so damaged that it couldn't find any sectors, certain sectors at all and then it would blissfully generate images. It was missing those sectors because it didn't find any evidence of them. Whereas if you tell it what the format is expected to be, it's supposed to be 18 sectors per track, then it knows to warn you if it only finds 17. Even if the other one, there's no evidence of the other one, it's been blasted from existence. A magnet being passed over or something.

Interviewer: Oh, that is interesting. Because I know IBM scans is used within the digital preservation community for like you say, it can tell you when it doesn't pick up certain sectors.

Interviewee: I think it's fine to use it with care. If you do a dump and then several tracks are a few sectors fewer than all the others, even without a warning, you'd probably be able to guess that there's something awry in there.

Interviewer: Which is good.

Interviewee: What I also want to do is add a feature where you can feed a dump and as I say, I've got this kind of these database or configuration files where there's a stanza per disk format. It says for each track or range of tracks, interleave and skew and so on, sector size. It should be possible to go from a good quality raw dump and automatically generate one of those. So I will add a feature to do that. 

Interviewer: Cool.

Interviewee: So you can say, here at least is a well-formed dump, generate that and then I've got a whole bunch more of those disks of unknown quality. But at least now I can say it's this format and now go and check them all against that. That'll be useful.

Interviewer: That would be very useful. I can see the community finding that really, really useful.

Interviewee: At least I envisage that it would be. This is the thing, it's like to some extent that's me coming up with a feature myself, which is always dangerous because they're not always the most useful features.

Interviewer: But that's a good one because that is something that does happen. We get, as you can probably imagine, quite a lot of the disks that we get in have not been looked after very well. So they've come from people's lofts or garden sheds, so mould or damage is sometimes found on them

Interviewee: Mould. So you must have a whole body of knowledge about cleaning disks.

Interviewer: I'm doing that with conservation next week which I'm really looking forward to because that is something that, sorry, a bit of a tangent here, but you've probably seen in a lot of forums, people discuss how to clean floppy disks, but there will be one person who says use soap and water and then the next person will say, do not use soap and water.

Interviewee: Don’t be, necessarily. Use, I don't know, lubricate the disk.

Interviewer: Then there's other people like don't lubricate your disk. 

Interviewee: Don't lubricate your disk. 

Interviewer: Don't do it because it's really bad for the drives. So it's very unclear what you should actually be doing. 

Interviewee: This work it’s really been scientifically tested.


It's a case of doing the gentlest cleaning you can without removing the coating off the plastic substrate. Because that's it, game over, isn't it? You can see through it, then you've gone too far.

Interviewer: But it's been quite fun because I've had people who've sent me mouldy floppy disks for testing and the library, as you can imagine, does not like it when you bring in anything that's mouldy.

Interviewee: Mould spores aren't that liked.

Interviewer: Not liked in libraries. So they're in a special room now. 

Interviewee: The mould room.

Interviewer: The mould room. The quarantine room. Which I think is very funny.

Interviewee: Have you seen some very dirty disks because I've had some pretty disgusting disks.

Interviewer: We've had a few dirty ones. I think I've been quite lucky, but I've seen some horrid examples within the community. I will actually send that to you as well. There's someone who last year at iPres, at that conference, did a talk on floppy disks that were found in someone's garage. They were not in good condition. But in the end, they did get clean reads out of nearly all of them after cleaning. So that is actually quite neat. So sometimes it looks really bad, but then it turns out to be okay.

Interviewee: Often mould seems to be quite cleanable. 

Interviewer: Which is good. I've also seen examples, I've not seen any of those myself, where the iron oxide layer is starting to … there was certain 5.25-inch ones that were bad.

Interviewee: Yes, this is where they recommend the lubricant, I think, so the drive doesn't scrape it.

Interviewer: So it kind of sticks on it.

Interviewee: So yes, it depends on the kind of damage you've got, right?

Interviewer: But just being able to know a little bit better what damage you're looking at is also really, really helpful. Let me just have a look if there's anything else specific that I wanted to talk about. Yes, actually a little bit about knowledge sharing and everything. Are there people that you would go to because I know that you talk about, of course on GitHub and everything, people they send you emails and request certain things. 

Interviewee: They do. 

Interviewer: But are there certain people that you would reach out to if there's something where you're like, oh, I'm struggling with this, and maybe it's good to have a chat with this person about that or?

Interviewee: Yes. Who would I talk to? There are other people doing similar software projects. So if I get someone who has an Apple II problem, it's not so much I'll ask, but I'll, or I might ask, but I'll also point them at the Applesauce Discord channel. Because they're very friendly there and there's a whole bunch of people there who know a lot about that particular domain. Or I might talk to the author of FluxEngine about a particular format and how you solve problems. It's kind of a small, or the Commodore 64, I've got a couple of people who at least have some knowledge about … they have some knowledge about the recording format used there. But it's a pretty small set of people. 

No, usually I'm working out by myself, I think is the short answer. But what's good is I like having the project. My forums are still on Facebook. I should move them to Discord probably, or something like that. Discord seems quite popular. But the GitHub ticketing and discussion areas are quite good because they do seem to be watched by people. So someone did for, I can't remember even what system it was now, some old 70s system. They came up with some dumps. They were like, can we work this out? I think there's some documentation here. Then someone else came along. 

Interviewer: Oh good.

Interviewee: Old, also interested in that system with more documentation. In fact, they even created an analyser and an additional patch to Greaseweazle to support it.

Interviewer: That's very cool.

Interviewee: Which I cleaned up and merged. So sometimes the experts come to you. The more open you are, the more likely that is to happen.

Interviewer: Because that GitHub, the questions section, I find very, very helpful. There's some very good examples or just people chatting about something where you're like, oh, that's probably what I'm also dealing with.

Interviewee: You can dive in. There's a whole bunch of people, so I think that's quite nice. I haven't experimented myself with Discord. Applesauce uses it. I think obviously Facebook put some people off. It's not the most kind of pick questions get asked. It's not a really great index of knowledge. It's hard to search or the search doesn't work very well. Then again, I'm not sure if Discord is great for that either. It's a bit of a stream of an internet chat.

Interviewer: It's kind of similar. I think it's difficult, isn't it? Because you just have groups of people…

Interviewee: People have tried to make new internet forums. What we need is a forum. It's like we probably do but unfortunately, if you just create one, then no one uses it.

Interviewer: Exactly. It's so difficult, isn't it, to find such things.

Interviewee: The kind of those GitHub Discussion is kind of a little bit in that direction, at least. But people do seem to actually use it, so.

Interviewer: It's nice that it's open, because with Facebook, of course and Discord's the same, it's more often not excluded because anyone can make an account, but it's more offline. Whereas if you were to look for a problem, from time to time a GitHub thread shows up, whereas that doesn't happen if it's on Facebook or Discord. Which is also a thing. Whatever you pick, there's always going to be drawbacks and benefits to them.

Interviewee: A bunch of people directly email me as well and that's always good. I don't get that many emails so I tolerate more to a degree.

Interviewer: Shouldn't say that, Keir.

Interviewee: Shouldn’t say that.

Interviewer: Which is very good. Is there anything else because I've gone through the questions I specifically wanted to ask you. Is there anything else that you think would be very useful to know or to chat about?

Interviewee: No, I think part of my interest in coming here today was to talk to you as a representative of the preservation community, because it's like I don't really have much knowledge of it and there's obvious scope for more conversation.

Interviewer: Yes, there is. That will be the first thing that I will do. I will probably get around to it on Monday morning, but I will send you a list of the resources and where we come from. As part of this project, one of the things that I will be developing with the community is more of a resource around floppy disk imaging and where to go to. So for example listing all the floppy controllers and why certain ones are what they're useful for for other things. So Applesauce, if you've only got Apple II disks, that's a good one. Like that but then there's others.

Interviewee: So you've got some of that sort of thing along those lines already.

Interviewer: So we did a workshop last year at iPres around obscure floppy formats and we wrote kind of a handout for people. It's only about 10 pages with a bunch of stuff. But that really took off. That was for me the moment where I was like, okay, hang on. There's more of an interest. It's not only us in a little niche part within a niche community. There's more interest in floppy disk imaging across our community and what best practise may entail. I always find that really difficult to say. I would say good practise instead of best practise. Like what is good?

Interviewee: Well, there's not a lot of documentation or not shared knowledge on that. In fact, this level of documentation for floppy disk dumping and preservation is very poor. 

Interviewer: Very, very poor. There's the KryoFlux Guide, The Archivist’s Guide to KryoFlux. 

Interviewee: That’s a guide.

Interviewer: So that one's quite good. But it's got some really good imaging and very basic set up stuff. But if you run into any type of problem or you can't get your hands on a KryoFlux, that already doesn't work anymore. I've chatted to the people who did the guide as well and they gave me some really good tips.

Interviewee: Flowcharts for the type of controller could be useful.

Interviewer: It's like oh, what do you have, that kind of stuff. You should go for this one. But also just guidance around where do you get a floppy disk drive? Who do you talk to? How do you make sure you get one that's actually a good, one that's useable?

Interviewee: These days even getting a good 5.25-inch drive is difficult. 

Interviewer: So difficult.

Interviewee: So difficult.

Interviewer: I've been very lucky. I put an email to Cambridge University, basically, and there were a whole bunch of people who were like, I've got one. So they're there behind you, in that cupboard.

Interviewee: Oh, look at that.

Interviewer: There’s quite a few of them.

Interviewee: You’ve got more 5.25-inch drives than I do. You've got a few 3-inch drives.

Interviewer: Yes, we've got 2 3-inch drives.

Interviewee: I've got more 3.5-inch drives. About 5 times that number.

Interviewer: Then there's some of the computers that have been donated.

Interviewee: Do you have an 8-inch drive?

Interviewer: No, we don’t have one of them.

Interviewee: No one in the university could even help you with that. 

Interviewer: No.

Interviewee: Or none that were prepared to give up. That’s the problem.

Interviewer: I've had a few emails from people who said, "I have this and this and this drive, but I'm not willing to share." Then it's a bit like, it's very cool that you've told me that you have it. 

Interviewee: What a moment. Wonderful for you.

Interviewer: So good for you but that's not really helpful. We don't have any 8-inch floppies in our collection and I'm not expecting them to be part of our collection just because most of the floppy disks come in with the personal archives and stuff. As you know, 8-inch drives or 8-inch disks were not really used for personal material. There's very small examples. So actually Churchill archives does have them because a politician used it for their agenda. 

Interviewee: What?

Interviewer: They were very early adapters. 

Interviewee: Very early.

Interviewer: So there are a small number of examples out there, but it's very, very limited.

Interviewee: Do you see hard sector disks at all?

Interviewer: I've not seen them in the wild.

Interviewee: They also seem quite rare actually, really.

Interviewer: I've had a few people that I've talked to who've had them. Actually both of them from a research institute,s which makes me wonder if it's also linked to … I don't actually know what was on their hard sector disks, but if it's linked to some type of research system or something. But I'm also expecting there to be some weird stuff that comes into the library just because Cambridge has such a rich history in computing. It's going to be weird systems all around. 

Interviewee: Oh yeah. Definitely. It has a long history of computing. 

Interviewer: Which is fun, but also again, that kind of worrying around documentation and making sure that we can actually make a copy that people can use. I'm excited, but also a bit nervous about it.

Interviewee: It's kind of a big space to chip away at. As you say, once you've got the good disk dump, it's what you do with it and accessing the data is...

Interviewer: It can be a real faff, but at least if we can get a good disk dump, it would mean that in a way, you could technically in a few decades…

Interviewee: Raw flux, multiple revolutions, three revolutions, don't skip tracks. You've got it there and you can work from that.

Interviewer: Why three actually, three revolutions?

Interviewee: It means that if it is marginal, at least you've got 3 tracks to compare across. This is also partially for copy protection and things. You've got this index mark or hole or whatever, or mark on the hub of a 3.5-inch disk, which marks kind of angular where the index point is on each track. Most systems will format a track from the index mark for revolution back to the index mark. So no data actually crosses the index mark. Actually for most systems, if you know you're getting a good dump doing one revolution, index to index would be enough. 

Interviewer: Oh cool.

Interviewee: However, not all systems do that. The Amiga didn't, for example, they ignored the index mark or some copy protections would deliberately ignore the index mark because then if you do an index to index copy, you put your right splice where you stop the right or what happened in the middle of the sector. Obviously therefore that sector will be bad. So then you need at least 2 index to index. If you're only dumbly doing index to index, you need 2 revolutions. 3 is like…

Interviewer: Just to be sure.

Interviewee: Just to be sure. It's not super.

Interviewer: But I think if there's nothing else there, I will stop recording. Thank you. Thank you so much for coming. That has been really, really helpful.

Interviewee: No problem.

[END]