On Wed, 2005-01-05 at 07:44 -0500, Dan Meltzer wrote: > 1. There is a fairly large amount of people who use gentoo on their > routers, a majority of these routers are pentium1/pentiumtwo boxes > with a processor speed of slow. These people > A. Do not want to use 2.6 because it takes up more space, and
"more space" with 2.6 is in the order of what? A megabyte once compiled? If these people are that low on space, they wouldn't be running Gentoo anyway with its enormous portage tree and its requirement of compiling software, which is also pretty space intensive. Were this a floppy-based router, I would agree with you completely. However, we are talking about a complete system that more than likely has a hard drive. After all, we were able to get the 2004.3 minimal LiveCD out in 50MB, and that is with a 2.6 kernel on-board. > B. Would be more concerned about the security auditing in 2.6 on a > router than they would on a regular box. I can understand this point to a degree. > However, To ask these people to do a stage1 install, which would take > them multiple weeks, does not sound like the best plan to me. We aren't building a GRP for them, so they would be building for a while anyway. There has to be a point when we stop and say "no more" otherwise we end up tied to a bunch of old software which none of us wants to maintain. By the way, I have a P166 machine running Gentoo. A stage1->working firewall took about 4 days, not multiple weeks. Nobody is forcing these people to use a 2005.0 LiveCD for their installation. We aren't destroying every copy of 2004.3 on all the mirrors and going commando to people's houses to destroy every copy in the world. People will still be able to do a stage3 from 2004.3 to get a completely 2.4-based system if they do not wish to run a 2.6-based system. The main point is that we don't want to build them anymore. So far, nobody has really given a compelling reason for us to add the extra work into building them. > 2. I believe Rac has made a post on the gentoo forums about how > stage1 is flawed, It does not add the original packages to world file > or something, and as such, he reccommends starting with a stage3 and > then then rebootstrapping and emerge world after starting a stage3. Well, here's my take on this. No bug has been filed on it... I have been looking into this, and plan on "fixing" the stage1 and stage2 targets in catalyst. I am actually a bit peeved at this situation, since -releng (who controls the stages and the stage building) has never been approached on this, but I'm leaving that one alone. Anyway, I am hoping to resolve that problem before the release. It means we'll have slightly bigger stages, but they'll be more accurate. > How would this work with a kernel switch in addition? Would it even be > possible/sensible to do? I'm not sure I understand this question. Could you rephrase it? > Two things I can forsee, not sure how they fit into your scheme of > things, but they might need to be looked at before releng goes totally > 2.6 Well, the second one really has nothing to do with switching to 2.6, at all. The first one is a valid concern, in part. However, I think we've pretty much agreed that we're not doing things just for the sake of saving a very small number of people some compiling time. Were we to be removing the option for 2.4, I could see an argument, but we are not. The user will have the same choice available to them, we just aren't doing all the work for them anymore if they chose to go with a 2.4-based system. This also brings up another question. Will the embedded team be switching to 2.6 for their profiles? People that are so concerned about disk space would probably do best to run the embedded stages for their architecture anyway in building their router. I know that the embedded team would love to get more users, as that is the exact target audience for their work. > On Tue, 04 Jan 2005 17:50:06 -0500, Chris Gianelloni > <[EMAIL PROTECTED]> wrote: > > On Tue, 2005-01-04 at 23:47 +0100, Francesco Riosa wrote: > > > >What the hell does having more than one kernel in /boot have to do with > > > >the LiveCD or creating stages? > > > > > > > >The kernel used on the LiveCD has exactly nil to do with what the user > > > >installs on his system. As a prime example, the 2004.3 LiveCD for x86 > > > >had only a 2.6/udev setup, but it didn't stop people from installing a > > > >2.4/devfs setup or a 2.6/devfs setup. > > > > > > > > > > > > > > > Can you forgot my question? > > > Symply I was thinking about stage3 and about kernel installed by stage3 > > > (instead the user install and configure it) > > > going home, I'm really sorry to have wasted your time > > > > This is completely off-topic now, but there's no way that we are forcing > > any kernel on anyone. Therefore, there will not ever be a kernel in any > > stage, as the stages are only used for building and are not supposed to > > represent a complete system. > > > > -- > > Chris Gianelloni > > Release Engineering - Operational/QA Manager > > Games - Developer > > Gentoo Linux > > > > > > > > -- > [email protected] mailing list > -- Chris Gianelloni Release Engineering - Operational/QA Manager Games - Developer Gentoo Linux
signature.asc
Description: This is a digitally signed message part
