I first became interested in using jigdo (Jigsaw Download) during the test releases for Fedora 8. I was spending a lot of time getting the DVD ISOs over bittorrent and was annoyed that in all probability I already had packages necessary to compose the release, but not in its official form from the rawhide trees I mirror locally.
Jigdo has the ability to start with a base image and build ISOs based on local (or remote) packages. I gave it try with some of the Fedora Unity Respins, helped test one of the Fedora 8 respins, and recently built the Fedora 9 Alpha ISO myself using jigdo.
There has been noticeably more interest in jigdo on the fedora-list@redhat.com mailing list over the past few months. This is interesting considering how summarily dismissed jigdo has been as “not being needed” by some of the regulars on the fedora-devel-list. The first time I heard jigdo mentioned was at the Red Hat Summit 2007 during the Revisor session. When I asked some of the Fedora veterans why Fedora didn’t use it I was told it was unreliable, full of memory leaks, and “that old thing from Debian”.
Is this the same disconnect between the folks who think CD releases of Fedora don’t matter and those that do? A classic “developer doesn’t think user wants or needs what the user says they need”? CD releases will be back for Fedora 9 and I think one of the reasons was the complaints from users on fedora-list.
The other arguments I’ve heard against doing jigdo is that it would not benefit enough users or would cause confusion by adding “yet another method to obtain Fedora”. I don’t find any these arguments strong enough to believe Fedora should not provide jigdo as a download option.
Looking at the Fedora Unity’s jigdo statistics people are using jigdo to get Fedora. How many more people could we get Fedora to if it was a mainstream option and how big is the risk if we try?
Speaking from personal experience I think providing direct jidgo support by Fedora would be a great thing. I have never had problems with jidgo or problems with the ISOs it creates. It provides a good alternative for people who cannot use bittorent or have limits on how much bandwidth they can use in a given period.
Who knows, maybe it will become a reality: http://fedoraproject.org/wiki/Features/JigdoRelease
March 31, 2008 at 8:47 am
Hi Jim,
it sounds like you did the right thing. The .jigdo file points to a a template needed to create the ISO you specify. It provides the basis for the ISO you will create locally. If you are having problems getting that file obviously the rest of the process will not work.
You might try viewing the .jigdo file and manually doing a wget to retrieve the respective template file and see if that works. Beyond that I’d ask the Unity folks for help. I’ve always found them to be particularly helpful and friendly. You’ll get the quickest help by contacting them on Freenode IRC on the #fedora-unity channel.
Hope that helps,
John
March 23, 2008 at 5:12 pm
I need help. I don’t seem to get the picture of using jigdo to download
the Fedora Unity respins.
I installed jigdo and jigdo-lite
I went to spins.fedoraunity.org/spins
I right clicked on Fedora Unity Re-Spin F8 20080204 and
saved the file, which is named Fedora-Unity-20080204-8.jigdo
I said jigdo-lite Fedora-Unity-20080204-8.jigdo
It gave me a list of 24 or so different images to choose from,
so I started with the first i386 CD image. It messed around for
some time and saved a file named something.template, then
said the checksum doesn’t match on the template.
The instructions seem to suggest that there is a separate .jigdo
file for each of the CD images, but I don’t know where those
files are.
February 21, 2008 at 12:53 pm
There’s a couple of issues with moving to jigdo as supporting both torrents and jigdo doesn’t make sense. Torrents also are completely community bound, as the community grows, so does the experience they have. The opposite is true with jigdo. We had a conversation about this back in december:
https://www.redhat.com/archives/fedora-infrastructure-list/2007-December/msg00018.html
Its worth it to note that the Unity guys are all for it (I would be too if resources for us and our mirrors were infinite) and in our case us hosting jigdo takes a huge burden away from the unity guys but puts it on us. Its a pretty complex topic to discuss with pros and cons in each. At the end of the day though, for Fedora and its mirrors, there’s a big cost upfront for the tiny portion of users that use jigdo right now. Could we change that by switching to jigdo? Possibly, we might also lose a bunch of mirrors in the process and cause the number of files we keep around to get much higher than we anticipate causing large costs.
We’re keeping an eye on this one though.
February 20, 2008 at 8:31 am
Any time I see one of Fedora’s Senior Leaders putting their “I’m going to pretend I’m just a user” hat on, it makes me very happy.
Marginal cost of making Jigdo work for Fedora? Very small. Fedora Unity has *already figured it out*.
Benefit — more happy users.
Great blog post, John.
February 20, 2008 at 7:12 am
This is typically a symptom of a developer having too much resource.
When CPU and memory on the development platform is too great, end-users see poor performance. When developers are sitting behind high bandwidth pipes, Jigdo is not needed.
Put developers on PIII 500MHz 128MB machines, and performance of the product will improve. Put developers on dial-up lines or 128Kbps DSL and Jigdo moves to the forefront.
Innovation is driven by the need to do more with less.
February 20, 2008 at 1:57 am
There’s no better example of why groups like Fedora Unity matter. A lot.
Engineers have their blind spots. People spend all day, every day, working on the project, and they develop biases. It’s just human nature, I think.