I was installing the OEM version of Windows 95 on VirtualBox the other day for reminiscence sake, and for some reason I had two different product keys. I tested both of them to see which one mapped to the CD, and to my surprise both product keys work with the installation. I always thought the product key and CD were one-to-one, but is this not so with Windows 95?

  • 2
    Welcome to Retrocomputing. This is a great question; it's already got an answer, which must be some kind of record here. It might be interesting to note that the same is true of Windows 98; when I lost the product key sticker it didn't take long to guess a working one. – wizzwizz4 Jun 7 '16 at 16:13
  • 1
    @wizzwizz4 I searched and didn't find a previous question like this one. Next time I will search harder. Thank you for your comment! – Retro Gamer Jun 7 '16 at 16:15
  • I wasn't saying it was a duplicate; it's not. I was commenting on the speed that it was answered, and that Windows 98 has the same behaviour. This really is a great question. – wizzwizz4 Jun 7 '16 at 16:18
  • 2
    It hurts me a little bit to see that the OS I grew up using is considered "retro" now. I'm only in my twenties! – user772 Jun 7 '16 at 18:33
  • 1
    Microsoft generally doesn't use a single key per CD. The key is generated by an algorithm specific to the product, and tied to a date-range (i.e. later builds won't accept older keys). The verification is built into the code (so a key for Professional Windows won't unlock a Home version, a key for Visual Studio won't unlock Office, etc). Site-license keys wouldn't work if it were unique. – peter ferrie Mar 1 '17 at 19:41

The CD media for OSR (OEM Service Releases) are generally all the same; there is no unique code on the disc. Imagine what a problem this would be for commercial users with multiple machines to support - they wouldn't want to archive multiple copies of the media.

Making the key unique to the disk doesn't add any protection if the disc image is easily writable (unless there is online checking to limit reuse of codes)

  • 8
    Also, production and damaged media replacements. If your original Windows CD was damaged/lost, you could order a replacement from Microsoft for (not-so-)token fee without obtaining a new license. – SF. Jun 7 '16 at 15:53
  • 5
    I can second this. I've been involved in installation of about 40 copies of Windows 98SE in one day. The license keys didn't give a hoot what CD was used to install from. Most of the CDs were left in their original packaging as backups. – Loren Pechtel Jun 8 '16 at 2:38

CDs are made by pressing, having a new master made is expensive so the content of the disk would rarely be changed, certainly not for a single copy.

Yes there are recordable CDs now but they were not common at the time windows 95 was released, cost more in bulk than pressing and generally have worse longevity.

Even for media that is not pressed making a load of identical copies is simpler than uniqely serialising disks.

I guess this was a large part of the motivation for having product keys in the first place, to allow a copy sold to be uniquely identified even though it was shipped on bulk-copied media.

  • So in a way, doesn't that mean that if you have a list of all the product keys, you could pirate windows and just use trial and error with the serial keys? – Retro Gamer Jun 7 '16 at 23:48
  • 11
    Sure you could, product keys are pretty ineffective at stopping people pirating software. – Peter Green Jun 7 '16 at 23:52
  • 1
    That makes sense. Of course nowadays at least with Microsoft, the serial keys use a database to limit the number of times you can use them. – Retro Gamer Jun 8 '16 at 0:01
  • @PeterGreen It doesn't seem like anything is pretty effective at stopping people pirating software. – I.Am.A.Guy Oct 5 '16 at 17:33

MSDN always uses the key 335-XXXXXX6 for legacy software which works for old visual studio & windows 95/98 editions. I just checked and VB6 still has the same key.

I don't think I ever remember typing other key back at the time (and don't forget internet access was a very new thing) I don't remember if it even came with a SLIP/PPP driver at the time.

  • Win95 definitely came with PPP bundled in, it was one of the first consumer OSes that made it easy to connect to the Internet out of the box for that reason. Windows 3.1 didn't include either. OS/2Warp3 only came bundled with SLIP and didn't have PPP included until Warp4 released in '96. I'm not sure when MacOS started shipping with PPP though. – mnem Jun 8 '16 at 18:03
  • I'm not so sure as I think you needed the plus pack to get IE see wikipedia here – PeterI Jun 9 '16 at 9:50
  • 2
    IE wasn't bundled with the original release, no. I don't think it was bundled until the later "B" release. But a SLIP/PPP dialer was included (unlike Win3.1 where you needed a third party one like Trumpet Winsock), which goes a long way towards making it easier to get connected for people not familiar with rolling their own networking. – mnem Jun 10 '16 at 2:42
  • 2
    Also don't forget that not including a browser wasn't a huge barrier at that time as you could walk into any store that sold software and buy a boxed copy of a browser of your choosing off the shelf. Plus I did say "connect to the Internet", not connect to the Web. There actually was a difference back then still. – mnem Jun 10 '16 at 2:50
  • I am 99% sure that the key you mentioned was not "the one key"; just the one distributed with MSDN. If you bought a product in retail, you would get a different key. Also, not every product used keys, and not all that did accepted that particular key; but many did. – Euro Micelli Mar 5 '17 at 22:57

Microsoft's licensing and activation functions have slowly evolved over time. When Windows 95 and then Windows 98 came out, Microsoft used keys that were unique but not tracked via a central database.

When Windows 95 was first introduced, at that time access to the Internet was still very uncommon, with people using 14.4 kilobit in about 1994 or so. All computers could not be guaranteed to have modems to call in to a licensing registry, and Ethernet was uncommon except in schools and businesses. So how would the software be able to easily know if a product key was being uniquely entered for each installation?


Typically in the old days, the way that install software would become activated involved using install media that could be written to. There would be some sort of key disk that would have activation information written to it, and was designed to use copy protection to prevent duplication and transfer of the software to someone else. If this disk was damaged or unreadable then then software could not be used anymore. Typically this type of software protection is only used infrequently.

If Microsoft had done this, the key disk would have had to be permanently inserted at all times, preventing use of the floppy drive, which would have been both cumbersome and annoying to Windows users.


The modern activation process is not specifically tied to the software installation disc, but rather to your hardware.

Most modern standardized Windows install media can be used an unlimited number of times, but requires a unique product key for each installation. The product keys themselves are tied to hardware inside the computer, commonly the MAC address of the network interface, which every modern computer now includes integrated by default.

For modern Windows installations that include Windows preinstalled and no product activation is needed, that is a special short-run customized installation media used by a vendor for all their products. That install media is not uniquely tied to that specific computer, but will work for any similar system from the same product family.

  • 1
    Disk keys were not required to be installed all the time for systems that needed them. The disk usually patched bytes on the primary application file that indicated the license and count. If you registered with the manufacturer via the pre-paid mailer the original registered owner could obtain replacement media. Netware was a major OS that used this method up until 4.2, after that it was just a printed key. – Rowan Hawkins Jan 5 '18 at 0:39
  • @RowanHawkins Another scheme I remember seeing back in the 5.25" floppy days was to combine what the 3.5" 720K disks for The Four Crystals of Trazere do (floppy-based copy protection for an installer that bakes a system fingerprint into the installed binary) with a token that could be moved back and forth between the installation media and the application with a special utility but not copied. – ssokolow Jan 13 at 0:25

CD distribution media is difficult/expensive to make unique. I once worked on a very low-volume product. We devised a technique to make the CD "proof of purchase" i.e. baked the product key into a unique CD which could not be copied.

What we did:

Our technique relied on burning a writable CD and then making a small scratch on the label side at a fairly specific radial distance. The CD distribution included one huge file which contained nothing more than the product key in a specific encoding and pattern. The scratch rendered some portion of this file unreadable. The installation process would read this file and expect to find it mostly readable with the correct data pattern, and some portion unreadable. If the file met this criteria, it would accept the CD and register the installed software with the license key encoded in the file. Any attempt to copy the CD would fail because of the damaged file.

What was wrong with our technique:

The technique itself worked well. We could reliably produce a CD that could not be copied and was proof of purchase. It worked for our low-volume product (a few dozen copies), might have been sustainable for hundreds of copies, but impractical for larger distributions because each CD had a one-off generated license file which took time to generate and even longer to burn the distribution onto the CD. Typical CD distributions involve a CD master which is pressed in the same way as vinyl records; each CD is identical.

Bottom line: while it is possible to bind a product key to a specific key (we demonstrated one way to do it), it isn't practical or economic to do for anything published in volume, especially not on the scale of an O/S.

  • 1
    This is a bit risky, as the top layer of a recordable CD tends to continue to delaminate once you've got it started. Especially if it's getting shoved in and out of one of those sleeve-books instead of a jewel case. Defendisk did something similar with 5.25" floppies back in the day, and even filed a patent on it back in 1984 (patents.google.com/patent/EP0129427A3). FWIW, this particular scheme is easily defeated by CD-ROM drive emulators, which generate errors for the blocks that failed to read. – fadden Nov 2 '19 at 20:47
  • 1
    @fadden I think the floppy case you cited is where our idea came from (it wasn't my idea). I know what you mean about delamination; we applied our own full-size label over top, so I don't think it would have been vulnerable in that way. Our product is now ancient history, so it no longer matters if the protection can be defeated. At the time, we deemed it secure enough. I was never in favor of distributing on physical media, preferring web-based download and activation back when most people still preferred to receive physical media. – Anthony X Nov 2 '19 at 21:16
  • Broderbund from the AppleII days did something similar. The fix for the the copy application was to just write an invalid checksum into the block. So even if the data were read supposedly fine, the block would return an error. – Rowan Hawkins Jan 16 at 22:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.