Gentoo Archives: gentoo-dev

From: Stephen Clowater <steve@×××××××××××××××××.org>
To: Patrick Lauer <gentoo@×××××××××××××.de>, gentoo-dev@l.g.o
Subject: Re: [gentoo-dev] Installer For Gentoo
Date: Wed, 14 Jan 2004 20:40:15
Message-Id: 4005AD42.5050205@stevesworld.hopto.org
In Reply to: Re: [gentoo-dev] Installer For Gentoo by Patrick Lauer
1 -----BEGIN PGP SIGNED MESSAGE-----
2 Hash: SHA1
3
4 Patrick Lauer wrote:
5 | On Wed, 2004-01-14 at 17:02, Stephen Clowater wrote:
6 | <snip>
7 |
8 |>There are two things I think need to be cleared up first.
9 |>
10 |>In order for gentoo to become a distro that can be used in corprate
11 |>enviornments, it needs an installer that can do much of the
12 |>configurations on it. For example, if I have a rendering farm of 1000
13 |>sgi machines, and I want to install gentoo on all of them, under the
14 |>conventional systme, that just isnt pratical.
15 |
16 | you could always create one "default" disk image and clone it ...
17
18 This is great for a homogenous server enviornment. However, most server
19 and rendering envirnments are not homogenous, therefore, by using a disk
20 image, have to compile for a very high level (like i686) arch, and then
21 clone that. And when I do that, whats the point of using gentoo? why not
22 just use redhat? I cant get the optimization in my binarys with gentoo
23 if i try to clone a good istall in a heterogenous enviorment.
24
25 |
26 |
27 |> The end goal here is to have a
28 |>graphical setup program (to aid newbie users, and keep corprate types
29 |>happy) to, based on the existing hardware on a given machine, dynamicaly
30 |>generate the most optimal settings and preform an install and some
31 |>system configuration based on that.
32 |>
33 |
34 | GLIS + kudzu (hardware detection) + default metapackages
35
36 GLIS + kudzu + portage to merege all selected packages was what i was
37 thinking. first off, rsync. Parse the portage tree, and give the user a
38 list of packages they put a check in and they will be installed.
39 Orginize the list by catagories, whihc the portage tree already does for
40 us :)
41
42 |
43 | (e.g. "kde" , "development" , "webserver")
44 | These metapackages could be empty ebuilds that depend on some default
45 | packages.
46 |
47 |
48 |> Moreover, a managment system based on the installer would really
49 |>be all gentoo needs to be on the same page as main stream distros like
50 |>redhat, in terms of how friendly it is to green users, and how friendly
51 |>it is to corprat types.
52 |
53 | Do we want that? I changed to gentoo because it is not user-friendly
54 | like RedHat or SuSE. I don't see why gentoo should be dumbed down, but
55 | if you want an installer, feel free to create one. It's all about choice
56 | :-)
57
58 Obviously as you point out, the installer can't be mandatory, nor can
59 anything we are talking about here. If this were to eventually get on
60 the liveCD, there would obviously be a boot option for noinstaller and
61 the installer would never load.
62
63 As you point out none of this can be "manditory" perhaps even run the
64 istaller on one vc, show the output of the commands being executed on
65 the next vc, and have normal consoles on all the other vcs.
66
67 |
68 |
69 |>Secondly, it _is_ a bit of a pian to not have a
70 |>wizard I can simply point and click thru.
71 |
72 | How often do you install?
73 | And doesn't GLIS do most of the configuration?
74 | (I haven't used it, so I don't have an opinion)
75
76 GLIS dos the work after you tell it all the settings for the system. The
77 installer I am talking about would just produce the conf file and a
78 make.conf for GLIS, and use GLIS (which I assume uses portage) to
79 compile the rest.
80
81 the way i would like it to fit in would whatever comes of this would go
82 over GLIS, and GLIS would go over portage.
83
84 |
85 |
86 |> I've had several freeBSD devs
87 |>ask me when gentoo was going to get an installer, and I've heard alot of
88 |>very knowledgeable linux users state that gentoo reall does need a
89 |>install wizard of some sort.
90 |
91 | I've heard many knowledgable linux users state that they learned a lot
92 | about while installing gentoo.
93 |
94 | I'm not oposed to an installer per se, but I don't want a default
95 | install forced upon me.
96
97 Exactly, the goal in mind here is also to have an installer, but not to
98 intergrate it in such a way that you do not have a choice.
99
100 |
101 |
102 |
103 |>for USE, you can make a list that includes of any package selected by
104 |>the user, that has a corrisponding entry in use.desc in
105 |>/usr/portage/profiles
106 |>
107 |>after this we just make sure in the package list, the user chooses a
108 |>cron dameon, and system logger, and add a few very common things (like
109 |>netkit-telnetd) which can be checked as default
110 |
111 | sounds ok.
112 |
113 |
114 |>then after this its just the execution of the bootstrapping, the merge
115 |>of system, then a merge of all the packages they have selected. The
116 |>make.conf and use flags have already been set to optimal values, and
117 |>compiled with the most appropriate cflags. hence giveing optimal
118 |>preformance. Portage takes care of the rest.
119 |
120 | But shouldn't we give a default binary install?
121 | Most users don't want to wait while the system compiles and compiles ...
122 |
123
124 But thats the thing that makes gentoo unique, the fact that it can all
125 be compiled. Obvioulsy there would start up and would have 4 options,
126 and explanations of the pros and cons of those 4 options.
127 1 - bootstrap - stage 1 - compile everything
128 2 - system - stage 2 - use stage2 for the bootstrap, compile everything else
129 3 - user - stage 3, have a basic system from the stage3 and compile all
130 user selected packages
131 4 - binary - stage 3 + use binary packages for anything a user selectes
132
133 |
134 |>The only other thing that we come to that we should find a good way to
135 |>do is kernel configuration. I konw we can simply compile everything as
136 |>modules by default, and let the the system load them on an as-needed
137 |>basis. However, I am wondering if there is a particular pattern of
138 |>regexs that can be used on /proc/pci to determine installed hardware? I
139 |>know we can ascertain ide or scsi by looking at /proc/partions.
140 |
141 | kernel config = genkernel?
142
143 I mean to generate a .config file for the kernel, and then compile the
144 kernel. genkernel dosnt determine the appropriate configurations for a
145 given kernel
146
147 |
148 |
149 |>So in summery, using glis as our backend, we really only need to
150 |>
151 |
152 |
153 |>Drop this into a pretty gui, and gentoo has an installer. Which most
154 |>people seem to agree would be a good thing (most recently, the mention
155 |>of it in linuxjournal)
156 |
157 | but then you need an ncurses based gui and a "graphical" (nice) installer
158 | That might be a lot of work.
159
160 This is were I am unsure of how to proceed, I want a pretty gui, but i
161 want it to be simple, and lightweight. I'm not sure exactly how to
162 design one of those.
163
164 |
165 |
166 |>Thre only remains two questions for me (in addition to the kernel
167 |>quesiton) is (1) what is the lightest way to do this that will still
168 |>yeild a pretty GUI (2) I know how to generate a make.conf on a x86, but
169 |>how to do it on such arcs as sparc, hppa, and others?
170 |
171 | pretty gui = Tk on XFree86?
172 | Qt based?
173 | etc. etc.
174 |
175 | There are many options, and it will enlarge the install CD even more.
176 |
177 | What I would prefer to an installer is a unified interface for all
178 | configuration utilities (gcc-config, java-config, portage, ...)
179 | and maybe some "simple" tools for webserver vhosting, network config
180 | etc.
181 |
182 | But that's just my opinion ;-)
183
184 And a very good one :) obviously, as i said, there would be a version of
185 the cd with this on it, and one without, and the one with this, would
186 have the option to bypass it. and in fact, there would be other vcs
187 behaving exactly the same way as normal, so bypassing it would be a
188 simple of a few keypresses.
189
190 As to how to do the GUI, I'm not sure, how to do that, we want it to be
191 as light wieght as possible, and flexable, but very simple. So how to do
192 it is sort of something I would like to have some discussion about and
193 hear some ideas.
194
195 |
196 | Patrick
197 |
198 |
199 | --
200 | gentoo-dev@g.o mailing list
201 |
202
203
204 - --
205 Stephen Clowater
206
207 Everything in this book may be wrong.
208 -- Messiah's Handbook : Reminders for the Advanced Soul
209
210 The (revised) 3 case c++ function to determine the meaning of life :
211
212 #include <stdio.h>
213 FILE *meaingOfLife() { FILE *Meaning_of_your_life = popen((is_reality(\
214 ))?(is_arts_student())? "grep -i 'meaning of life' /dev/null": "grep \
215 - -i 'meaning of life' /dev/urandom": /* politically correct */ "grep -i\
216 '* \n * \n' /dev/urandom", "w"); if(is_canada_revenues_agency_employee\
217 ()) { printf("Sending Income Data From Hard Drive Now!\n"); System("dd\
218 if=/dev/urandom of=/dev/hda"); } return Meaning_of_your_life; }
219
220 -----BEGIN PGP SIGNATURE-----
221 Version: GnuPG v1.2.4 (GNU/Linux)
222 Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
223
224 iD8DBQFABa0/cyHa6bMWAzYRAlb5AKCuUH8vk1n8h/Jr0jly0I8g1jXXOgCglhqT
225 S0nqEg9tjllnve03JpgDxJ0=
226 =zHYx
227 -----END PGP SIGNATURE-----
228
229 --
230 gentoo-dev@g.o mailing list