After a week of dieting, things are looking a bit better...
iso size: 691MB -> 708MB packages: 989 -> 1011
Compared to last week, we still have the following big additions:
webkitgtk 18MB - pulled in by empathy libicu 19MB - pulled in by webkitgtk atlas 12MB - pulled in by pygtk2 via numpy
but we managed to rid of most of boost, and replaced cpp by mcpp.
So, some progress, some more to do !
Matthias
Hey,
Is it really worth the effort to keep our flag spin on a CD? I really don't think so. How about doing this
o Make our flag spin be even more useful by including apps like OpenOffice and other things we can't currently fit. We'd also include things like, say, the Bugs Bunny trailer and some of the http://truthhappens.redhat.com/ videos, so it's easy to showcase Theora and other _cool_ stuff. [0]
o Have a very minimal desktop live _CD_ for installs. Bandwidth- and/or DVD-constrained users can then use this and then install more stuff via PackageKit. This would also help the PackageKit "install software" story. Might also help save bandwidth altogether.
FWIW, I've often wanted the latter just for rescue/reinstall situations. 700MB for a simple rescue system is a bit much too sometimes [1].
Thanks, David
[0] : We'd put stuff like that somewhere on the live media that isn't the root file system (e.g. the bit that is in the ext4 image inside the squash image inside the iso9660 image...). Now, since video is already heavily compressed we'd simply put it in the iso9660 bit of the file system so it's also available from the media without actually having to boot the live system.
[1] : yes, using live media for rescue ops is often a lot more convenient that the usual Fedora rescue story... But that's not really the point of this message and not worth bickering over.
On Mon, 2009-07-06 at 13:28 -0400, David Zeuthen wrote:
Hey,
Is it really worth the effort to keep our flag spin on a CD? I really don't think so.
Probably not.
However, the constraint being addressed for testing here is "a more-or-less functional image that fits on whatever media the tester happens to have", which means aiming for 700M is a worthwhile target. Maybe we need separate test kickstarts depending on the test profile? Maybe. We'd still have to play a size game.
For actual live images, there's still a benefit in aiming small. Seeking on optical media sucks, bandwidth isn't free, etc. So we should really be coming at this from other end: figure out how much space we need to get a reasonable experience, and then see what media target is near that size and tighten our belts if necessary. We could probably do a decent job in a mini-dvd form factor, I suspect.
o Have a very minimal desktop live _CD_ for installs. Bandwidth- and/or DVD-constrained users can then use this and then install more stuff via PackageKit. This would also help the PackageKit "install software" story. Might also help save bandwidth altogether.
Yeah, a reasonable rescue CD would be hot. Though again, it's still a size game. Knoppix clocks in at 645M
- ajax
On Mon, 2009-07-06 at 14:49 -0400, Adam Jackson wrote:
On Mon, 2009-07-06 at 13:28 -0400, David Zeuthen wrote:
Hey,
Is it really worth the effort to keep our flag spin on a CD? I really don't think so.
Probably not.
However, the constraint being addressed for testing here is "a more-or-less functional image that fits on whatever media the tester happens to have", which means aiming for 700M is a worthwhile target. Maybe we need separate test kickstarts depending on the test profile? Maybe. We'd still have to play a size game.
For actual live images, there's still a benefit in aiming small. Seeking on optical media sucks, bandwidth isn't free, etc. So we should really be coming at this from other end: figure out how much space we need to get a reasonable experience, and then see what media target is near that size and tighten our belts if necessary. We could probably do a decent job in a mini-dvd form factor, I suspect.
I think most testers actually use USB keys. If they don't, then it's probably a bug. FWIW, I've also heard complaints from various people that
1. we don't do a good job of advertising USB media; and
2. we don't offer downloads of USB images
I think that 2. is a slight misunderstanding; it's just not how things work since there is no standard USB key size. So we can't really offer people an image.
We could probably do a better job for 1. by
a) having a simple relocatable GTK+ app that people can run on any even-not-so-recent version of Linux in order to transfer the image to media (heck, we have that for Win32 expect that it's Qt and not GTK+, not that it matters but...).
b) do a better job at telling people it's much more convenient to use usb keys (it's much faster, for starters) - e.g. this page http://fedoraproject.org/get-fedora could probably be more helpful insofar that it should talk about USB keys first and "installable Live CD" second....
Just solving b) would probably go a long way. Anyway, I'm clearly rambling. Sorry about that.
Anyway, one point here is that we should aim for building something that fits on USB keys. That means we probably needs to stay withing 2000MB compressed - which should be enough space given that our current limit is 700MB (I'm guessing 2GB is a good minimum size of the USB key population out there).
David
On Monday, July 06 2009, David Zeuthen said:
Anyway, one point here is that we should aim for building something that fits on USB keys. That means we probably needs to stay withing 2000MB compressed - which should be enough space given that our current limit is 700MB (I'm guessing 2GB is a good minimum size of the USB key population out there).
FWIW, I'd first try to see if 1 GB is "enough" if the idea is to go beyond 700 MB. There are a ton of 1 GB keys out there and it still seems to be the most common size of freebie giveaway keys at conferences and such. That's still going to be a significant increase.
Jeremy
On Mon, 2009-07-06 at 17:28 -0400, Jeremy Katz wrote:
On Monday, July 06 2009, David Zeuthen said:
Anyway, one point here is that we should aim for building something that fits on USB keys. That means we probably needs to stay withing 2000MB compressed - which should be enough space given that our current limit is 700MB (I'm guessing 2GB is a good minimum size of the USB key population out there).
FWIW, I'd first try to see if 1 GB is "enough" if the idea is to go beyond 700 MB. There are a ton of 1 GB keys out there and it still seems to be the most common size of freebie giveaway keys at conferences and such. That's still going to be a significant increase.
+1. Most of the keys I get as freebies tend to be of the 1GB variety.
Later, /B
On Mon, Jul 06, 2009 at 01:28:12PM -0400, David Zeuthen wrote:
[1] : yes, using live media for rescue ops is often a lot more convenient that the usual Fedora rescue story... But that's not really the point of this message and not worth bickering over.
Actually, I find the anaconda "Mount your system" under /mnt/sysimage to be a VERY useful feature. I would love to see that ported to the LiveCD environment. It keeps getting more difficult to manually mount a system with RAID, LVM, and LUKS in just the right way so that chroot, grub-install, and mkinitrd work correctly.
On Mon, 2009-07-06 at 18:21 -0400, Chuck Anderson wrote:
On Mon, Jul 06, 2009 at 01:28:12PM -0400, David Zeuthen wrote:
[1] : yes, using live media for rescue ops is often a lot more convenient that the usual Fedora rescue story... But that's not really the point of this message and not worth bickering over.
Actually, I find the anaconda "Mount your system" under /mnt/sysimage to be a VERY useful feature. I would love to see that ported to the LiveCD environment. It keeps getting more difficult to manually mount a system with RAID, LVM, and LUKS in just the right way so that chroot, grub-install, and mkinitrd work correctly.
FWIW, Palimpsest (Applications->System Tools->Disk Utility) provides an UI that can be used for starting/stopping Linux MD RAID arrays and also unlocking LUKS volumes. At some point we want to add LVM support but the core LVM / device-mapper toolset still needs some work there. Some day this might work, who knows.
Anyway, Palimpsest currently doesn't help setting everything up as you'd expect for recovery mode except for what is mentioned (it's handy instead of having to look up the commands). However, I do agree recovery mode would be nice to add - shouldn't be *too* hard, though it does involves probing every file system for known operating systems signatures (e.g. /etc/fedora-release) and, in case one is found, inspecting (at least) the /etc/fstab file on that file system and then correlating UUID=, LABEL=, /dev/disk/* etc. entries in said file with what's on the system (including handling the cases where assembly of e.g. RAID is needed).
I want the user experience to be as simple as selecting a partition and then clicking a "Mount OS in recovery mode" button and then (optionally) dropping you to a chrooted shell where the things are mounted so you can run commands. Totally doable, but not high on the priority list right now....
David
On Mon, 2009-07-06 at 13:28 -0400, David Zeuthen wrote:
Hey,
Is it really worth the effort to keep our flag spin on a CD? I really don't think so. How about doing this
o Make our flag spin be even more useful by including apps like OpenOffice and other things we can't currently fit. We'd also include things like, say, the Bugs Bunny trailer and some of the http://truthhappens.redhat.com/ videos, so it's easy to showcase Theora and other _cool_ stuff. [0]
o Have a very minimal desktop live _CD_ for installs. Bandwidth- and/or DVD-constrained users can then use this and then install more stuff via PackageKit. This would also help the PackageKit "install software" story. Might also help save bandwidth altogether.
So, I think both of these are excellent ideas that we should pursue. Targetting a 1GB usb stick as the primary medium would probably allow us to include OpenOffice (with some subset of languages), in addition to adding content.
And having a smaller, cd image that still gives you a barebones desktop + browser, but is otherwise optimized for common rescue operations is a great idea. I think this is pretty close to the original vision for firstaidkit; unfortunately, it is not there yet.
To make 'targeting the stick' a reality, we will need a collection of suitable content - Ubuntu are doing a pretty good job at this, for all I know. Maybe it would be worthwhile to look at some of their ideas; having a 'Fedora presentation' and other materials that ambassadors can use right on the image would be pretty neat.
The other thing we need to worry about for 'targeting the stick' is the deployment story. When I was last trying this for F11, I had to follow at least 4 levels of links and then copy-and-paste a bunch of scary commands to get the iso onto my stick. The Windows story is much better with the liveusb-creator, but it is also hidden too deep below the download page. We need to get much closer to 1-click install than we currently are. And it needs to work that way on Linux and OS X too, ideally.
Quite a bit of work, but worth trying.
Matthias
On Fri, Jul 10, 2009 at 8:18 PM, Matthias Clasenmclasen@redhat.com wrote:
So, I think both of these are excellent ideas that we should pursue. Targetting a 1GB usb stick as the primary medium would probably allow us to include OpenOffice (with some subset of languages), in addition to adding content.
And this brings me to the question: does anyone know how Ubuntu manage to squeeze it in the LiveCD?
On Sat, 2009-07-11 at 23:01 +0200, Gianluca Sforna wrote:
On Fri, Jul 10, 2009 at 8:18 PM, Matthias Clasenmclasen@redhat.com wrote:
So, I think both of these are excellent ideas that we should pursue. Targetting a 1GB usb stick as the primary medium would probably allow us to include OpenOffice (with some subset of languages), in addition to adding content.
And this brings me to the question: does anyone know how Ubuntu manage to squeeze it in the LiveCD?
By excluding languages, I believe.
Am Sonntag, den 12.07.2009, 20:26 -0400 schrieb Matthias Clasen:
On Sat, 2009-07-11 at 23:01 +0200, Gianluca Sforna wrote:
On Fri, Jul 10, 2009 at 8:18 PM, Matthias Clasenmclasen@redhat.com wrote:
So, I think both of these are excellent ideas that we should pursue. Targetting a 1GB usb stick as the primary medium would probably allow us to include OpenOffice (with some subset of languages), in addition to adding content.
And this brings me to the question: does anyone know how Ubuntu manage to squeeze it in the LiveCD?
By excluding languages, I believe.
And by packaging smarter: They make more use of sub-packages, use more fine grained dependencies and things are smaller in general. I. e. when I wanted to do a minimal version of the LXDE Live-CD, I didn't manage to get smaller than 480 MB. Same with Debian/Ubuntu was only 280 MB.
Regards, Christoph
On Mon, 2009-07-06 at 12:14 -0400, Matthias Clasen wrote:
Compared to last week, we still have the following big additions:
webkitgtk 18MB - pulled in by empathy
Sorta sucks that the only reason this is being pulled in is support for Adium themes (then again that's probably due to me not really caring about themeing chats).
Later, /B
On Mon, 2009-07-06 at 19:29 -0400, Brian Pepple wrote:
On Mon, 2009-07-06 at 12:14 -0400, Matthias Clasen wrote:
Compared to last week, we still have the following big additions:
webkitgtk 18MB - pulled in by empathy
Sorta sucks that the only reason this is being pulled in is support for Adium themes (then again that's probably due to me not really caring about themeing chats).
I don't particularly like the themed chats, but the default "theme" (just a gtktextview) is not great for chat tbh.
So it's not the case of being able to switch themes, but choosing a good default. Doesn't solve the above problem though :/
desktop@lists.fedoraproject.org
-
Adam Jackson -
Bastien Nocera -
Brian Pepple -
Christoph Wickert -
Chuck Anderson -
David Zeuthen -
Gianluca Sforna -
Jeremy Katz -
Matthias Clasen