Windows 95 in 4.47MB

Silvorgold writes "BOFH of MSBetas.net has been able to compress Windows 95 into 4.47 megabytes, making it the world's first sub-5mb bootable, registry editable, command-promptable, usable version of Windows 95. He has written a small description about what he did, and also included screenshots (with his digital camera), and don't worry, these aren't fake screenshots."

Comments

Re:Why?

[PurpleFloyd]

Two reasons I can see:

1. Because it's there. While Linux is fairly easy to get a useful Linux distro under 2 MB, you can do things like strip the kernel to the bare essentials. Needless to say, you can't do that under Windows; there's a lot more challenge in getting Win 95 under 5 MB.
2. Also, making Win95 fit in small spaces may be of interest to people who want to run legacy Windows apps on embedded devices. I could see this put on an old Pentium with an all-in-one motherboard and a 16 MB solid state drive, with room for a small program or two. The only issue would be swap space. This might be useful in places looking for a small, simple pseudo-embedded PC that needs to run Windows apps. Linux might be better for 95% of these kinds of tasks, but if Windows is necessary to run legacy apps, then it would be best to run, well, Windows.

While it might not be eminently practical, neither is, say, running Linux or NetBSD on some obscure piece of hardware. Nifty hacks like this aren't always done for practical reasons; they're just as often done for fun.

But 5.35 MB version used no executable packer?

[The Revolutionary]

From this page, it appears that the previous 'record holder', 5.35 MB, did not use an executable packer or other compression.

"Apparently only 5.35Mb in size (at the moment.... I'm told this might go down!) - without using UPX / any compression"

So, is what this fellow has done a superior acheivement, or did he mostly just run an executable packer on a few binaries?

Certainly if the idea here is to just shrink the physical disk space usage we can do better than either of these entries by compressing all files and hacking the Windows I/O subsystem calls to handle our compression.

I think all of this raises an interesting question. (ok, so it's not interesting at all, but I've had similar issues come up in a lot of other unofficial sort of 'competitions' like this, and we all just kind of use interest at that point ;). Just what is the purpose of this, and at what point do your modifications, whether extreme, or just running binaries through an executable packer, defeat the purpose of doing this in the first place?

Is the idea to have the smallest possible OS capable of doing x or y?

Is the idea to have the smallest possible OS that looks like Windows 95?

Is the idea to have the smallest possible 'distribution' of Windows 95 attainable by just removing unecessary features?

Do we want smallest in terms of RAM usage, or smallest in terms of disk space? What do we then if we run it on a RAM disk? Which space counts?

Surely depending up just what is the goal here, we can do a lot better than 4.47 MB.

I guess I don't 'get it', what they're doing =)

That's Windows users for you!

There is a micro Linux distribution floating around somewhere that provides an X server in under 2 MB of physical disk space (but 4 or 8 MB of RAM), but I can't recall the name of it just now.