NB: At some point I started using ChatGPT to help me navigate the technical issues around getting metadata for a ripped CD on Windows 98, so the entire chat might be a useful archive of that. (assuming open.ai is still around in 20 years). What’s more, I had it draft an original version of this post which I entirely re-wrote, but as I continue to play with the machine for writing, I think it is only right to be completely transparent about the process.
I talked about finding a Minidisc player a few posts ago, and I spent some serious time over the last week or two getting this thing working (in every sense of that word).
Let me start with the basics. When I got it in 2018 or so the unit would not power on, so I packed it away for several years until recently. The “gumstick” battery had corroded, so I needed to order a replacement which I eventually did. In the interim the unit had a 3V DC port I could use to see if it had life when powered on. It did and the unit works perfectly, which was nice to learn. Once the battery came I tested that out, but unfortunately the contacts were corroded all to hell.
The above video from sjm4306 takes you through cleaning up the corrosion, recommending white vinegar and alcohol—which both worked well. I also incorporated his recommendation to thread a wire on the battery door to ensure there’s consistent contact. With that fix the player now reliably powers on and holds a consistent charge with the battery.
Windows 98 machine in the bavacrypt where the minisdisc experimenting happens
That was the easy part, what follows is pure retro-computing masochism. I spent way too much time trying to get this unit working cleanly on an original Windows 98 machine in the bavacrypt. As mentioned in the previous post on the minidisc, one of the things that got me excited about the project was the ability to connect a Minidisc to a computer and transfer files using the old school Sony SonicStage software (now discontinued). It’s also always fun to connect peripheral hardware to an old school computer and remember how things used to work.
The above video takes you through the process of getting SonicStage working with a Minidsic player seamlessly. The main difference for me was that he’s using a virtual Windows XP machine to make things cleaner when transferring files you want to load on the minidisc. In the road less travelled part in the Minidisc world, I decided to install SonicStage (version 3.0) on my Windows 98 machine and try the process with the OG Pentium III machine. SonicStage installed without issue, and I was able to cleanly transfer music files from the computer onto the disc, it was all so simple. If you like simple stories, stop here cause everything that follows is designed to complicate all this intolerable simplicity.
After getting it to work, I was thinking: “now comes the fun part of ripping my CD collection so I can start taking the long road back to physical/virtual media that I actually own.” Upon putting a CD into the computer and importing it into SonicStage I assumed metadata like tracks, album, and artist info would just appear automagically (as they say). Turns out, that’s not the case. I have a fuzzy recollection of this—probably because it was all happening seamlessly behind the scenes—but audio software like WinAMP and SonicStage were dependent on the Compact Disc Database (CDDB) to retrieve all this metadata because CDs don’t have this info. Turns out a CD contains almost no human-readable information. There’s no artist field, no album title, no track names. It’s just a table of contents: how many tracks there are, how long each one is, and where they start and stop. The WAV format is pretty much a black box.
Now, here come all the complications. Turns out the Windows 98 machine connected to the internet in the bavacrypt is not exactly a web-first machine. Https has made browsing the modern web next to impossible from an old machine. I do have the protweb.org proxy running so I can browse webpages archived through the Wayback Machine cleanly on a 1999 version of Internet Explorer—there’s already a post about this rabbit hole. But what I quickly realized when importing a CD into SonicStage is that the software is depending on a call out to a long defunct Gracenote owned CDDB server to grab all the disc’s metadata. For years this worked well enough that nobody thought about it. You put in a CD, names appeared, and you moved on. The fact that this required a live Internet service, maintained by a company, with rules and permissions and quotas, was easy to ignore until you try and live like it’s 2003 on the modern web—retro-computing is nothing else if not a lesson on how ephemeral and fragile it all can be.
Knowing I had to redirect the CDDB lookup to a service that does work, namely Gnudb,* the trick was finding a way to have SonicStage call the Gnudb service at gnudb.gnudb.org:80/~cddb/cddb.cgi so that I could use this new server over port 80 (sans https). Long story short, it doesn’t work. I edited hosts files. I rewrote registry keys. I forwarded ports. I even made a Windows 10 machine pretend to be the Internet in 2002 and point everything to gnudb.org. Nothing. SonicStage doesn’t just “look up CD info,” it identifies itself with a hardcoded client ID, talks to specific servers, and assumes those servers still care that SonicStage exists. They don’t. Sony and Gracenote shut down support for these older clients years ago, not accidentally but deliberately—don’t trust Sony or Gracenote to keep the web whole for us retro-computing geeks.
So, admitting defeat, I moved on to Winamp. At first I thought I cracked the code when it happily announced “CDDB: Success” for every disc I inserted. I was naively thinking the proxy on the Windows 10 machine was working, although I wasn’t seeing any metadata added to the tracks. Turns out it announced success even when I unplugged the Ethernet cable. WTF? It did it after I deleted its cache. It did it when I pointed it at servers that didn’t exist. Like SonicStage, there was no place I could find in Winamp 2.95 to change server names and point to gnudb.org. Eventually I learned that for this version of Winamp “success” doesn’t mean “metadata found.” It doesn’t even mean “network access occurred.” It means “the CD drive successfully read the table of contents.” That’s it. Winamp considers this a win and sees no reason to complicate things further. But we both know I do.
I spent a lot of precious time swapping CD input plugins, because different versions of Winamp shipped with different ideas about how CDDB should work. Some had hardcoded servers. Some allowed editing. Some hid everything in INI files. Some looked editable but weren’t. It started to feel more like an archaeology dig than anything else. To be fair, part of this project was an excuse to put AI Maddeness on ice for a week or two, but it was starting to get ridiculous. In case you lost the thread, all we’re trying to accomplish here is to get artist, album, and song metadata for a physical CD ripped on a Windows 98 machine—why does everything have to be so hard?
Eventually I moved on to yet another audio software of the era: Exact Audio Copy (EAC). Turns out the third time truly is a charm. EAC used another service (also now defunct) to grab disc information called Freedb, which forked off the CDDB project in 2001 when Gracenote bought CDDB and it turned its back on the community that created it—turns out enshittification is an age-old story on the web.† EAC version 0.95 (not 0.99 as I learned the hard way) works cleanly with Windows 98 and it was dead simple to replace the Freedb server settings with the Gnudb server settings. While holding my breath, I asked EAC to look for disc info and it actually worked. REDEMPTION! Three days of work and finally a little love for Windows 98.
Freedb server settings that I could re-write to point to gnudb.gbudb.org
Now that I figured out how to grab the metadata remotely, now I had to figure out how to apply it regularly in EAC so it’s recognizable in SonicStage. [Shakes fists at the sky!] While EAC 0.95 can fetch artist and album information just fine from Gnudb, WAV files themselves are a terrible place to store that information. Seems the WAV format was never designed to carry tags in a standardized, universally respected way. There are multiple hacks and conventions to try and deal with this, but SonicStage mostly ignores all of them. Folder names don’t reliably help. Filenames don’t reliably help. You can do everything “right” and still end up with a library full of anonymous tracks once you import them.
This is where the old advice starts to collide with present reality. For years, the standard wisdom was to use WAV with SonicStage. The reasoning was simple and technically “sound:” CD audio is lossless PCM, WAV is lossless PCM, and ATRAC (the proprietary codec MiniDisc uses) is lossy. CD to WAV to ATRAC means one lossy encode. CD to MP3 to ATRAC means two. Therefore, WAV is technically “better.” That said, SonicStage doesn’t manage WAV metadata very well, probably relying heavily on CDDB to fill in the gaps. With CDDB defunct, WAV stopped being friendly. MP3, on the other hand, works beautifully. SonicStage reads ID3 tags reliably and consistently. Artist, album, track names all come through without complaint.
So you end up with a wannabe purists dilemma: is avoiding a second lossy encode actually worth manually typing metadata for every album? On Minidisc hardware, at Minidisc bitrates, for portable listening, the honest answer is probably no. Most people cannot reliably hear the difference, and many of the people who say they can are probably lying, or simply remembering a time when the distinction felt more important than it does now.
SonicStage was built for a world where CDDB worked, WAV was temporary, and metadata lived inside proprietary libraries that no one imagined would be defunct twenty years later. Once those assumptions collapse, the “best” workflow becomes the one that causes the least friction. For some people, that still means WAV, manual metadata entry, and the satisfaction of knowing there was only one lossy step. For others, it means using MP3 as a metadata carrier, letting SonicStage read what it understands, converting to ATRAC, and moving on with their lives. While neither choice is wrong, one is purer and the other is kinder 🙂
I decided to use EAC to create both WAV and compressed MP3 files simultaneously using the compression tool lame.exe, which is something we needed for Audacity back when we did audio in ds106. It brought me back. So now the ripping takes twice as long and I get the original WAV file and a compressed MP3 file (sometimes as much as 1/10th the WAV). Not only does the MP3 codec handle metadata well, but it does a damn good job at compression which explains why it won out in the end.
What all this rear-view mirror navel gazing made clear for me is that a niche audio device connected to the web 25 years ago tells a familiar story. Namely what started as an open project, namely the CDDB, quickly moved from a community-driven tool to a corporate-owned convenience that just happened to be a key piece of a fairly fragile system. As the rise of online music made CDs more and more obsolete CDDB vanished and everything downstream started to feel brittle. Software didn’t so much break as it outlived the world it was designed for and was abandoned. The only reason this service is still alive is because it was originally written under a GNU license and folks could and did fork the code to route around the obstacles corporations created in the name of growth/profit.
As we learned from Goodfellas, once they’ve busted it out, they “torched the place.” There’s no sense of responsibility or stewardship once the value has been extracted, and to be fair I’ve seen the same approach play out in more than a few large-scale web projects at universities as well — so it’s not just evil corporations. The web is too often treated as disposable infrastructure, and that’s unsettling when you consider how much of our cultural memory now lives there. Turning things off and walking away is easy; reckoning with what’s been lost is harder. This is why entities like the Internet Archive, and passion projects like GNUDB, are doing what feels like God’s work — not preserving nostalgia, but refusing to let the web become a memory hole. More of us need to be doing that.
*Thanks to the passion and good will of the project maintainer there’s an alternative that doesn’t depend upon giving over your data or cuddling up with the corporate behemoths. A very cool project that I’ve now benefitted from directly in my strange misadventure through the world of music CD metadata. I just wanted to give a huge thanks the ther person behind Gnudb, and a donation to follow cause I know what it means to try and keep the lights on. Also, if of any interest, I run a hosting company and we do have some server space and CPU we can donate to the project if that might help.
†Gnudb is the last database standing in this lineage, if you will.
