One of my duties at Apress is to commission books on Linux. I also author Linux books myself. My job involves keeping a close eye on the whole Linux distro scene. This in turn involves downloading many ISO images so I can install and evaluate various Linux distros.

Playing around with Linux all day isn’t a bad job to have but it’s hampered by the fact I have the world’s slowest DSL connection. OK, I exaggerate. At 512Kbps it’s one of the slowest broadband connections available where I live. This is because of my distance from the nearest telephone switch.

60KB/sec makes downloading ISOs tricky. CD ISOs usually arrive intact, free of corruption. With DVD ISOs I have a roughly 50/50 hit/miss rate. And, frankly, I’ve had enough.

The last DVD ISO I tried to download was Mandrake 2007. I wanted to write a review of it on this blog, in fact. So I started the download early on Friday. I left my office computer switched on over the weekend, to download the ISO, and had to check every few hours to make sure the download hadn’t stalled (I work from home).

Real life got in the way of my checking and, sure enough, the download stalled several times and for several hours, meaning that it took until Sunday evening for the download to finish. At that point I checked the md5sum and, yup, it was wrong. I decided to burn it anyway but there must have been significant corruption because the ISO wasn’t recognized and wouldn’t burn.

Arrgghh! Must… suppress… murderous… impulse…

Even if only one byte in the entire file is wrong, the md5sum will be different and I have no way of knowing if there’s serious corruption involved. In fact, I once burned a SUSE Linux DVD ISO that had a bad md5sum and it transpired one package file was corrupted. By careful package selection I could work around it and still use the DVD.

There has to be a better way of distributing Linux ISOs than this. Linux is all about accessibility and availability, but by relying on ISO images as one of the main distribution mediums, we’re effectively limiting Linux to those who have fast Internet connections. To the rest we’re saying they can take their best shot at downloading but success isn’t guaranteed.

What I’d like to see is some system of managed downloads, maybe where separate smaller chunks are downloaded and checksummed there and then. If there’s corruption then the chunk is downloaded afresh. Almost certainly, the solution is to download smaller files—it’s when betting the house on one large file that the risk of corruption is unacceptably high.

Some distros have FTP-based installers, whereby packages are downloaded on demand. I guess this is a step in the right direction, although in my experience FTP-based installs have always involved complicated setup (more complicated than a straight install, anyway).

Maybe a scheme like I’ve described already exists, although I suspect an entirely new file transfer protocol is called for. Let me know in the comments below if you know of one. In fact, let me know what you think of this whole issue, because I think this is a problem that needs fixing, especially as ISO images get larger and larger. Let’s try and get some momentum on this issue.