RSS 
There has been some talk about taking some of the current direction of Gentoo and running with it, mainly, turning Gentoo in to a massive collection of overlays. Part of me sees this as appealing, and on further reflection I think I know why. It presents it self kind of like dystopic sci-fi does, which is kind of cool and appeals to me. But part of me is against it as well. Fractured environments can be cool and in various fields yield to crazy cool thing like evolution from competition and stuff, but on the flip side, you don't get much tight integration or cross communication, which is what most people are saying is the biggest problem with this plan, or more specifically, lack of ability to handle inter-dependencies. I think it could be a bigger problem than that, but we'll get into that presently.
Something to keep in mind with Gentoo directional discussions is that we may all want different things from our distros. I'm going to discuss what I want. Others may disagree because they want different thing's from their distro. That's super cool in fact and one of the reasons I like Gentoo is that it seems to accommodate more people better than other distros. Moving on.
Interestingly enough I'd still be excited to see seeds take off and get things like a Gnome seed, but I'm not that enthused about a Gnome overlay which then has Gnome removed from Portage proper. Maybe examining why will highlight the differences in the plans and show what I find myself looking for in a distro.
One of the reasons I'm pondering Ubuntu is the integration between packages at different levels of the stack and integration on the whole. For instance, your basic Ubuntu install now comes with a nice splash boot screen. But hark you say, Gentoo can do that, and so it can. Why Gentoo even supports two (last I checked) different splash screen setups. Of course your default system doesn't have either. You have to tack that on after you do a base install.
Another example is getting suspend working on laptops. Again, Gentoo supports a few ways (suspend in kernel proper and suspend from the suspend sources from what I remember). Again, neither are available by default but with some work you can get one up and running. I'm using suspend sources for example. So all is good, no? Well, unfortunately, higher up the stack, gnome-power-manager is having some weird problem (for me at least) that once it wakes up from a hibernate it doesn't reconnect to the hardware and battery. So in some situations if I go into hibernation with little battery life and come out plugged in and charge my battery up gnome-power-manager still acts like I have no charge and am on battery, not AC, and so tries to go back into hibernation when I'm not looking. It's a pain. And I suspect Ubuntu wouldn't ship with this kind of bug.
I love Gentoo for giving me so much power and some many options, don't get me wrong, but with that power comes some responsibility. A Gentoo system by nature requires a lot more love (read: time) to do the things other systems can do by default. On the flip side, if you happen to not like some other systems defaults you may just be out of luck, where as with Gentoo, if you put in some effort, you can make it do anything.
So Gentoo vs. Ubuntu to me seems to be a a trade of Power and customizability vs. ease of use in certain areas. A Gentoo system needs more love/time, and an Ubuntu system is more tightly integrated and needs less love/time, but also has less options.
As for seeds vs. overlay, lets look at it. I don't see moving Gnome, KDE and webapps into separate overlays as being conductive to the problems I've outlined above with my desktop. If anything it should further fragment the layers yielding less integration. But if seeds work like I hope they will, we should have the same level of integration we have today, but also with a tool for installing a massive set of preset packages and configurations for them that work together and have been tested together yielding integration across a stack and operating more tightly together, more like Ubuntu. With seeds I see it as possible to get the ease of management of Ubuntu and awesomeness of it's tested integration in areas but still keep Gentoo's customizability and power overall which I love.
So what it seems to come down to is that on a gut level and from certain view points, the overlays approach is cool, but the seeds approach can give us better integration. And these days I am looking for better integration from my distro. I want it to work well together across many levels and I think seeds is born to do that while it looks live overlays will not address this at best and quite possibly further exacerbate the problem.
Note 1: I very much love the overlays for things like testing experimental software before importing it into portage proper. In this post I'm simply arguing against fragmenting the portage tree as it is into a bunch of overlays.
Note 2: I don't just use Gentoo on my desktop, I also use it on my server and love that apache in Gentoo/portage has support for mpm_peruser, because I think with almost any other distro I'd have to roll my own install to get mpm_peruser and nothing would probably go as well, Gentoo currently does work quite well as a server, that's part of it's awesome flexibility.
