Gentoo Archives: gentoo-dev

From: Kent Fredric <kentnl@g.o>
To: gentoo-dev@l.g.o
Subject: Easy Installs / Stage 4 ( Was: Re: [gentoo-dev] Re: Packages up for grabs )
Date: Sun, 07 Aug 2016 16:56:46
Message-Id: 20160808045555.639b6770@katipo2.lan
In Reply to: Re: [gentoo-dev] Re: Packages up for grabs by james
1 Moving this to a higher visibility thread because I don't know how many
2 people think such an intense discussion is happening in the tail of a
3 package assignment ..... :P
4
5 Most of my negativity/limitations are more about making sure we define
6 where we aught to go.
7
8 They're less "Limits", more "guides", often enough.
9
10 On Sun, 7 Aug 2016 12:24:37 -0500
11 james <garftd@×××××××.net> wrote:
12 > Sure, I agree here, but, statistically these "hi level" languages are
13 > being taught, in lieu of C; and that is really sad. I'm sure there
14 > are exceptions, would you have a few CS departments that push C over
15 > java and the other, newer languages? (I'm curios).
16
17
18 Here, the "Limitation" as such is that once you realise that "C" is not
19 the only target, nor is Java, you realise there are going to be end
20 users who want "easy mode" and have no intent on doing any immediate
21 development work.
22
23 Or we might have user bases who want ready-to-go media for python
24 development.
25
26 This is not a "limitation", because if you think about customization,
27 we can very much provide all the choices.
28
29 And theoretically, we can provide named sets of default good choices
30 that are "reasonably good for the purpose they describe", and then
31 people can pick and choose what they want, and then if they want to
32 expert-mode it, they can hand craft their own "choice set", or progress
33 to installing their base system like the rest of us.
34
35 >
36 > >
37 > > You can't satisfy everyone out of the box.
38 > >
39 > >
40 > > The rest of your response kinda rotates around a central axiom that
41 > > makes other Linux distributions effective, and "Easy":
42 > >
43 > > The lack of choice, a tailored work flow, a target audience, and a
44 > > narrow focus on what the vendor delivers.
45 > >
46 > > Gentoo is fundamentally unlike these things, because the Gentoo way
47 > > has always been first and foremost about *user choice* and
48 > > *maximising user choice*
49 >
50 > Noting in promoting an easy install semantic for a default, buildable
51 > system, precludes choice after the system is installed and boot. For
52 > examle a default install, using Calamares and ending up with KDE,
53 > could easily then have kde removed and lxqt installed. That would be
54 > up to the new user to figure out, via the handbook and the wiki....
55 >
56 >
57 > > The reality is a giant hunk of the world are *not interested in
58 > > choice* They want something that works and get out of their way.
59 >
60 > Quite true, but we're talking about increasing gentoo's update
61 > amongst those linux leaners, not converting windows/mac users that
62 > are not interested in alternatives....
63 >
64 > >
65 > > That's why proprietary systems with deep, vertical architecture and
66 > > product lock-in are still incredibly popular.
67 > > They understand their market, and they focus on making things work
68 > > for that market by tailoring it to a very narrow set of features
69 > > that satisfies 95% of its target.
70 >
71 > Support is always a crowd pleaser, imho. So with fresh ideas, the
72 > newest members support those right behind them in line with user
73 > level issues. Noobs helping noobs. buntu has proven this works, if
74 > nothing else.
75 >
76 > >
77 > > Gentoo's target audience is decidedly that other 5%, the group of
78 > > people who don't mind getting their hands dirty, the group who wade
79 > > up to their elbows dealing with horrible problems because that's the
80 > > consequence of the power of choice.
81 >
82 > What I proposed, models for easier install and a VM/Container system
83 > that is secure and allow for experimentation with "jentoo" does not
84 > limit, but, encourage choice and experimentation.
85 >
86 > Let's focus on the easy install. Once folks get a running gentoo
87 > system, most figure out how to manage it and like the choices, build
88 > from sources (and bin packages for the larger/complex).
89 >
90 > >
91 > > You can promote pre-boxed Gentoo products if you want, I just think
92 > > you're barking up the wrong tree if you think doing that will help
93 > > anybody.
94 >
95 > You are misconstruing the message. It's a boxed, quick install that
96 > would behave going forward, with the same (exact) semantics as a
97 > grudge-filled traditional install. The only difference is that first
98 > install is
99 > quick, fast and easy. Nothing else changes, unless this fresh install
100 > chooses to embrace additional packaging or alternative packages
101 > compare to the default install. Nobody needs to make that decision.
102 > Surely many will then go read the handbook and the wiki to move
103 > forward. The install just becomes painless for a few basic or default
104 > examples. We do currently provide an occasional 'liveDVD'. So just
105 > image one of those, with an easy install pathway.
106 >
107 > >
108 > > As with most open source, it requires volunteer effort to make this
109 > > happen, and its a hard sell to try to convince existing staff to
110 > > spend more time on producing a thing that exists only to *reduce*
111 > > user choice for the sake of convenience.
112 >
113 > Again, you are incorrectly suggesting that these easy installs will
114 > preclude traditional gentoo semantics for adding, modifying, patching,
115 > or any other form of currently available modifications or enhancement
116 > from occurring. I'm not certain if you are twisting the focus here
117 > intentionally, or you are just limited in your imagination? Nobody
118 > wants that (artificial) limitation, so why would it me the semantic
119 > going forward, after an easy install?
120 >
121 > Think of it like sex. All of the traditional would be wonderfully
122 > available, but we're just adding a quickie (install) as an extra
123 > option. No limitations, just *choice* on the install.
124 >
125 >
126 >
127 > > And I just think most of our devs have more interesting problems to
128 > > solve than that, and you'd be simply weakening the core Gentoo
129 > > development team by trying to steal existing Gentoo staff to
130 > > engineer this carefully designed and polished "Just Works For
131 > > Noobs" platform.
132 >
133 > Agreed. My idea is some encouragement and maybe receive a little bit
134 > of positive advise. The noobs will help the noobs, and a few will
135 > migrate down the maintainer--> dev pathway.
136 >
137 > On this list and elsewhere gentoo devs have admitted to using quickie
138 > installs, and liked it. It's just frowned upon to document it and
139 > encourage it. Like a wiki page on how to convert a calculate or
140 > sabayon install to a traditional gentoo system.
141 > >
142 > > And even then, I think if you did OK, it would be striving for the
143 > > wrong thing.
144 >
145 > An easy install, does not have to be detrimental. Over the years I
146 > have taught quite a few 'youngsters' how to drive on rural flat land
147 > in a big 4x4 with an automatic transmission and a booster seat. You
148 > just put the transfer case into low, and they cannot go very fast and
149 > the love the *power*, spinning tires and slinging mud and riding
150 > around. Later on in life they all have matured into productive adults.
151 >
152 > Face it, gentoo is a power trip, we all know that. Letting folks feel
153 > that, in a easy, but real, quick install default version, that
154 > eventually hooks them into gentoo.
155 >
156 >
157 > >
158 > > If you're going to come to a competition that has existing major
159 > > players ( such as the "noob friendly" linux desktop market ), you
160 > > have to not be simply a "me too!" in order to hope for success.
161 >
162 > The power of Gentoo-traditional awaits them, soon after reading and
163 > learning from the handbook and the wiki. Not all will use this, but,
164 > it sure would put a more attractive face on taking gentoo for a test
165 > drive.
166 >
167 > >
168 > > You have to have something unique that blows all the competition
169 > > out of the water in at least one way, that capitalises on an
170 > > un-tapped need.
171 > >
172 > > Anything else will just be some pathetic copy-cat attempt.
173 > >
174 > > And for Gentoo, our "Unique Edge" *is* our configurability, our
175 > > incredibly effective and convenient flexibility.
176 >
177 > Again, nothing in this idea inhibits the full power of gentoo from
178 > being available; nothing. The begging of (their) journey is easier and
179 > more appealing. Many will stay in the valley of the noobs, but other
180 > will turn to the handbook and the wiki; just as grasshopper became
181 > shaolin, imho. Monk my words.....
182 >
183 >
184 > >
185 > > Sacrificing our primary benefit to chase after some other market
186 > > half-assedly ... I can't see that panning out well myself.
187 >
188 > You keep with this false choice, that is non-sequitur. The only
189 > limiting here is your mind.
190 >
191 >
192 > >
193 > > Personally, I think we need to double down on what we're good at,
194 > > flexibility, and configurability.
195 >
196 > No arguments there. Putting an experimental for of gentoo, complete
197 > with questionable java, into a secure Gentoo hosted VM or Container,
198 > is not flexibility and configurability?
199 >
200 >
201 > >
202 > > Find ways of building systems at the users behest that do exactly
203 > > what they want easily, and not assume we know what is best for our
204 > > users.
205 >
206 > You should go to any of the progressive job boards (stackoverflow
207 > etc). Java is everywhere. It should not be mandated but a choice.
208 > Sincere the are numerous issues java, secure it via a VM or a
209 > container, if that can be done?
210 >
211 >
212 >
213 > >
214 > > Anything else and Gentoo will go in the direction of the sad sorry
215 > > state of the Linux Desktop, where neither GTK/Gnome or QT/KDE are
216 > > very useable anymore, and they've become encumbered with horribly
217 > > lethargic and bloated design, because they were all trying too hard
218 > > to chase what they thought people wanted, the standard established
219 > > by Windows and OSX for "Easy".
220 > >
221 > Agreed and that is not what I'm proposing, either. I'm proposing an
222 > easy install for a few types of basic systems (that choice is open to
223 > discussion). And, if possible, a way to put either a secure VM or a
224 > secure container on a hardened gentoo system to put an
225 > insecure/experimental form of jentoo into.
226
227
228 Here is my incomplete idea of how I would see this working out in our
229 favour.
230
231 For one, some context: On Quora, I frequently see questions in the form
232 "I have <blah> condition and <blah> requirements, what is the right
233 linux for me and my hardware <blah>"
234
235 There are literally so many such questions its mind-numbing.
236
237 Explaining the nuance and benefits of lots of different alternatives is
238 a pain in the ass, and it would be nicer to simply say "Gentoo" for
239 each and every question, then tack onto it "Under configuration X"
240
241
242 Here's what I think would be "Nice".
243
244 1. "Nice" Installer that did the grunt work, but which grunt work it did
245 was not actually specified by that installer.
246
247 2. Instead, the "Nice" installer is fed a recipie file, and the recipie
248 file glues user input and system configuration together, which the
249 installer than "Makes happen".
250
251 3. Recipie files would in part contain a SHA1 containing the git commit
252 on repo/gentoo.git that it used to install from, and they would end up
253 at the end of their execution provisioning /etc/portage and friends,
254 and finialising an install.
255
256 4. The key part of relying on "Set of specified configuration and SHA1"
257 is repeatability.
258
259 Historically a lot of the headaches I've found with Gentoo was you
260 tried to install your brand new system, and oops, the configuration you
261 chose ran into some bug and you had to think about things, because the
262 tree got updated between the time the instructions were written and the
263 time you tried to do them.
264
265 And that kind of problem is bound to leak into any installer that
266 installs "Fresh" from updated sources.
267
268 This means with a "recipie + SHA1" combination you can either
269
270 a) Download the recipie only and bootstrap that recipie into a full
271 system downloading everything as you go
272
273 b) Use that recipie to pre-build resources and assemble installation
274 media that allows you to perform an identical implementation of a, sans
275 internet access.
276
277 Though doing both of these would be hard in some cases, because if the
278 recipie is configuration-free in regards to compile flags, then both
279 can be done. But if configuration needs to change compilation, 'b' gets
280 a little trickier.
281
282 Essentially, I'm kinda thinking beyond "Just a thing" and thinking in
283 the direction more of "A meta thing".
284
285 If you want to make people feel powerful, how much more can you get
286 than doing what we've done with .ebuilds?
287
288 Ebuild being bash may be a horrible format, but man, every day I use it
289 and see what everyone else is doing, I just love it. You can get insane
290 loads of stuff done with incredibly little effort.
291
292 And if you want to give people power trips, what could be more powerful
293 than empowering everyone to create their own flavour of
294 gentoo-based linux just by tweaking a bash script?
295
296 Then we'd just have a maintained repository of these recipies that are
297 vetted by devs and signed.
298
299 And you could even have a net-inst installer that starts out by syncing
300 that repository and giving you a list of recipies to install from.
301
302 Emerge yourself a whole gentoo :P
303
304 NB: I'm really tired, so I'm in that "can't tell if this is brilliance
305 or deranged rambling" mode.

Replies