Mailing List Archive

Marijn Schouten (hkBst) wrote:
> I was wondering what is keeping us from releasing a minimal install cd more often than we do now.
> Isn't almost everything needed for it already in the stable tree and thus tested? And if so isn't
> it possible to fully automate generation of these cd's?

Time and sanity. Sure, it can be built pretty quickly, but there's a bit of
testing that needs to go into it before it can be released to the mirrors.
Release time is hectic enough, and you want us (releng) to do it more often than
bi-annually?

Also, what's the point? Everything you use to install from the minimal is
fetched from the internet. The only thing that an updated minimal would give us
is a slightly more hardware support during the install.

--
Andrew Gaffney http://dev.gentoo.org/~agaffney/
Gentoo Linux Developer Installer Project
--
gentoo-dev@gentoo.org mailing list

Marijn Schouten (hkBst) wrote:
> I was wondering what is keeping us from releasing a minimal install cd more often than we do now.
> Isn't almost everything needed for it already in the stable tree and thus tested? And if so isn't
> it possible to fully automate generation of these cd's?

I'd rather like to see that the stages are being updated more often.

Cheers,
Tiziano

Tiziano Mueller wrote:
> Marijn Schouten (hkBst) wrote:
>> I was wondering what is keeping us from releasing a minimal install cd more often than we do now.
>> Isn't almost everything needed for it already in the stable tree and thus tested? And if so isn't
>> it possible to fully automate generation of these cd's?
>
> I'd rather like to see that the stages are being updated more often.

This would certainly make more sense than updating the minimal CD, but the same
reasons for not doing apply as for the minimal.

--
Andrew Gaffney http://dev.gentoo.org/~agaffney/
Gentoo Linux Developer Installer Project
--
gentoo-dev@gentoo.org mailing list

On Thu, Mar 01, 2007 at 11:49:25AM +0100, Marijn Schouten (hkBst) wrote:
> I was wondering what is keeping us from releasing a minimal install cd more often than we do now.
> Isn't almost everything needed for it already in the stable tree and thus tested? And if so isn't
> it possible to fully automate generation of these cd's?
As agaffney points out, it's more that our actual CD releases are
intended to be well tested. While we could do daily builds of the
install CD, you really would not gain much beyond minor hardware fixes.

As opposed to minimal CDs, I would see weekly builds of stage3 tarballs
to be much more useful.

--
Robin Hugh Johnson
Gentoo Linux Developer
E-Mail : robbat2@gentoo.org
GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85

Hello everybody :)

I'd like to argue about this problem, since my point of view slightly
differs about this "problem".

I think that releasing installation media more often would make sense,
more than in the case of stages. I wrote a small article a few weeks ago,
about the "Small Gentoo" LiveCD (1). In just a few days, it has become one
of the most read articles of my site. People really often encounter
problems with recent hardware, and I must confess that I had sometimes to
bring them to "alternative" LiveCDs (2), because Gentoo minimal lacked
such hw support.

One way could be to automate LiveCD generation, as Debian or Ubuntu do
with their "weekly /nightly builds", but not to provide support on it.
Maybe we could have two kinds of releases : experimental releases,
automatically generated, and "official" releases ?

About the stages, I think things are well as they already are. Upgrading
to "latest gentoo release" isn't really a big problem, when you have the
hardware support ("emerge --sync ; emerge system -e" solves this
problem most of the time ?).

Regards,

Hubert.

(1) https://forums.gentoo.org/viewtopic-t-505165.html
(2) http://www.sysresccd.org/
--
gentoo-dev@gentoo.org mailing list

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Andrew Gaffney wrote:
> Also, what's the point? Everything you use to install from the minimal
> is fetched from the internet. The only thing that an updated minimal
> would give us is a slightly more hardware support during the install.

This slightly more hardware support is almost always what new boxen need. Of course people can
install Gentoo from another more up to date other distro, but wouldn't it be better if that were
optional and not mandatory? That if people have new hardware which is not supported by their own
distro, that they can pop in the latest gentoo cd and know that if there is any distro whose
install cd supports their hardware, that gentoo will too?

There would also be many more chances to fix things, since these images would be relatiely short
lived. Since each image would be more similar than the previous one than our current 6 months apart
releases are to eachother, testing could be spread out more. And our users would get more chances
to help us test. Releng might not have to take a snapshot and try to stable it a la debian. Instead
all developers could work together to fix bugs in stable found by users testing the install cd.

Marijn


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFF5t+5p/VmCx0OL2wRAnoBAJ93c8wOmeVYVtDuLwVO5Qwly9sNogCfbPON
7Leo1TTCqecCdo3sFJ0huRY=
=tcD4
-----END PGP SIGNATURE-----
--
gentoo-dev@gentoo.org mailing list

I personnally would like to see stage tarballs updated more frequently
if an arch receives a major update in system, like gcc or glibc, even
if this is only an r patch because these are a pain to install, and by
update I mean something new goes stable. Such releases are infrequent,
but make it painful when doing a new install because you know that you
have to rebuild world from the start.

On 3/1/07, Marijn Schouten (hkBst) <hkBst@gentoo.org> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Andrew Gaffney wrote:
> > Also, what's the point? Everything you use to install from the minimal
> > is fetched from the internet. The only thing that an updated minimal
> > would give us is a slightly more hardware support during the install.
>
> This slightly more hardware support is almost always what new boxen need. Of course people can
> install Gentoo from another more up to date other distro, but wouldn't it be better if that were
> optional and not mandatory? That if people have new hardware which is not supported by their own
> distro, that they can pop in the latest gentoo cd and know that if there is any distro whose
> install cd supports their hardware, that gentoo will too?
>
> There would also be many more chances to fix things, since these images would be relatiely short
> lived. Since each image would be more similar than the previous one than our current 6 months apart
> releases are to eachother, testing could be spread out more. And our users would get more chances
> to help us test. Releng might not have to take a snapshot and try to stable it a la debian. Instead
> all developers could work together to fix bugs in stable found by users testing the install cd.
>
> Marijn
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2.0.2 (GNU/Linux)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFF5t+5p/VmCx0OL2wRAnoBAJ93c8wOmeVYVtDuLwVO5Qwly9sNogCfbPON
> 7Leo1TTCqecCdo3sFJ0huRY=
> =tcD4
> -----END PGP SIGNATURE-----
> --
> gentoo-dev@gentoo.org mailing list
>
>
--
gentoo-dev@gentoo.org mailing list

Marijn Schouten (hkBst) wrote:
> This slightly more hardware support is almost always what new boxen need. Of course people can
> install Gentoo from another more up to date other distro, but wouldn't it be better if that were
> optional and not mandatory? That if people have new hardware which is not supported by their own
> distro, that they can pop in the latest gentoo cd and know that if there is any distro whose
> install cd supports their hardware, that gentoo will too?

Sure it would be nice, but this has its price too, namely a lot more
time spent on bringing releases out.

> There would also be many more chances to fix things, since these images would be relatiely short
> lived. Since each image would be more similar than the previous one than our current 6 months apart
> releases are to eachother, testing could be spread out more. And our users would get more chances

Uhm, just because you release more often, that doesn't mean you need
less testing to get the same level of quality.

That being said, I think this is really up to the releng team and noone
else. They are doing the work, so we can discuss it far and wide, as
long as releng doesn't want to do it, nothing will happen. So maybe we
should wait for a statement from Chris before doing anything else.

--
Kind Regards,

Simon Stelling
Gentoo/AMD64 developer
--
gentoo-dev@gentoo.org mailing list

Simon Stelling wrote:
> That being said, I think this is really up to the releng team and noone
> else. They are doing the work, so we can discuss it far and wide, as
> long as releng doesn't want to do it, nothing will happen. So maybe we
> should wait for a statement from Chris before doing anything else.

I can tell you right now what Chris's answer is going to be.

If you're volunteering to do it, join releng and have at it. As it is now, most
members of releng do not have the time and/or desire to do a release more often
than bi-annually.

--
Andrew Gaffney http://dev.gentoo.org/~agaffney/
Gentoo Linux Developer Installer Project
--
gentoo-dev@gentoo.org mailing list

I just want to point out that I have gone through 4 years of
gcc/glibc/binutils/baselayout/udev/etc. updates on 8 production servers
with varying hardware _with the hardened profile_ and I've never had to
re-emerge world once.

In addition, I don't see a huge drawback to using other distro's LiveCDs.
Some distros specialize in making a LiveCD do everything possible, like
Knoppix. This group isn't maintaining a portage tree, they are just making
their LiveCD awesome. Why not piggyback on their efforts (our users are
doing it anyway)? That way we can focus our resources on making portage
great, which is Gentoo's true strength.

With Gentoo, once you are up and running, releases become very
unimportant. What do you think?

-Cory

On Thu, Mar 01, 2007 at 09:22:34AM -0500, Caleb Cushing wrote:
> I personnally would like to see stage tarballs updated more frequently
> if an arch receives a major update in system, like gcc or glibc, even
> if this is only an r patch because these are a pain to install, and by
> update I mean something new goes stable. Such releases are infrequent,
> but make it painful when doing a new install because you know that you
> have to rebuild world from the start.
>
> On 3/1/07, Marijn Schouten (hkBst) <hkBst@gentoo.org> wrote:
> >
> >Andrew Gaffney wrote:
> >> Also, what's the point? Everything you use to install from the minimal
> >> is fetched from the internet. The only thing that an updated minimal
> >> would give us is a slightly more hardware support during the install.
> >
> >This slightly more hardware support is almost always what new boxen need.
> >Of course people can
> >install Gentoo from another more up to date other distro, but wouldn't it
> >be better if that were
> >optional and not mandatory? That if people have new hardware which is not
> >supported by their own
> >distro, that they can pop in the latest gentoo cd and know that if there
> >is any distro whose
> >install cd supports their hardware, that gentoo will too?
> >
> >There would also be many more chances to fix things, since these images
> >would be relatiely short
> >lived. Since each image would be more similar than the previous one than
> >our current 6 months apart
> >releases are to eachother, testing could be spread out more. And our users
> >would get more chances
> >to help us test. Releng might not have to take a snapshot and try to
> >stable it a la debian. Instead
> >all developers could work together to fix bugs in stable found by users
> >testing the install cd.
> >
> >Marijn
--
gentoo-dev@gentoo.org mailing list

Cory Visi <merlin@gentoo.org> wrote:
> In addition, I don't see a huge drawback to using other distro's LiveCDs.
> Some distros specialize in making a LiveCD do everything possible, like
> Knoppix. This group isn't maintaining a portage tree, they are just making
> their LiveCD awesome. Why not piggyback on their efforts (our users are
> doing it anyway)? That way we can focus our resources on making portage
> great, which is Gentoo's true strength.

++ on that. While LiveCDs draw new users due to spiffy things you can do with
"your" GentooCD, i'd like the releng folks to concentrate to make a stable and
well tested release (which collides with the nightly proposal). To top it off:
nightmorph was quite right in pointing out, that we need to accompany these
releases with documentation -- gentoo in my eyes draws greatly in this two
areas: flexiblest package manager around *and* greatest docu around.
We should keep it that way.
--
Kind Regards, Matti Bickel
Homepage: http://www.rateu.de
Encrypted/Signed Email preferred

> Simon Stelling wrote:
>> That being said, I think this is really up to the releng team and noone
>> else. They are doing the work, so we can discuss it far and wide, as
>> long as releng doesn't want to do it, nothing will happen. So maybe we
>> should wait for a statement from Chris before doing anything else.
>
> I can tell you right now what Chris's answer is going to be.
>
> If you're volunteering to do it, join releng and have at it. As it is now,
> most
> members of releng do not have the time and/or desire to do a release more
> often
> than bi-annually.

Or better yet just grab the specs and a beefy machine and build the stuff
yourself...that is what it's there for.

Build the automation stuff, bribe solar for a blade to run it on. Run it
for a month, then just tell releng "Hi I can build liveCDs and stages
every 3 days, look at my progress over the last month". Everyone likes to
see work, not 'oh lets do this' followed by nothingness :P

As always, talk is cheap.



--
gentoo-dev@gentoo.org mailing list

Robin H. Johnson wrote:
> As opposed to minimal CDs, I would see weekly builds of stage3 tarballs
> to be much more useful.
>

++ on that. Rebuilding _everything_ because openssl changed ABI since
the last stage3 install could save users a lot of time/trouble. (that
was just an example)

Weekly, bi-weekly or even monthly stages feel enough for me :)

Those could be released as "experimental stages".

Rémi
--
gentoo-dev@gentoo.org mailing list

Rémi Cardona wrote:
> Robin H. Johnson wrote:
>> As opposed to minimal CDs, I would see weekly builds of stage3 tarballs
>> to be much more useful.
>>
>
> ++ on that. Rebuilding _everything_ because openssl changed ABI since
> the last stage3 install could save users a lot of time/trouble. (that
> was just an example)
>
> Weekly, bi-weekly or even monthly stages feel enough for me :)
>
> Those could be released as "experimental stages".

...which people will still bitch and moan about when broken. People seem to have
an unreasonably expectation for everything we release, even if it's
"experimental", to work perfectly.

--
Andrew Gaffney http://dev.gentoo.org/~agaffney/
Gentoo Linux Developer Installer Project
--
gentoo-dev@gentoo.org mailing list

Cory Visi wrote:
> In addition, I don't see a huge drawback to using other distro's LiveCDs.

I tried to install a system with a marvell pata drive, no livecd I know
of managed to run because they couldn't find the cdrom. the system was
to be installed using debian and had windows installed, the debian.exe
got really handy (btw what about having one for gentoo or a generic one?)

> Some distros specialize in making a LiveCD do everything possible, like
> Knoppix. This group isn't maintaining a portage tree, they are just making
> their LiveCD awesome. Why not piggyback on their efforts (our users are
> doing it anyway)? That way we can focus our resources on making portage
> great, which is Gentoo's true strength.

well gentoo livecd usually run on many more systems, our infrastructure
let us have fresher stuff than others. a 2007.0 snapshot would have
worked out of box.

>
> With Gentoo, once you are up and running, releases become very
> unimportant. What do you think?
>

that first you have to be up and running... so release early & often
minimal livecd please =P

lu

--

Luca Barbato

Gentoo/linux Gentoo/PPC
http://dev.gentoo.org/~lu_zero

--
gentoo-dev@gentoo.org mailing list

On 3/1/07, Cory Visi <merlin@gentoo.org> wrote:
> With Gentoo, once you are up and running, releases become very
> unimportant. What do you think?

That's true, but ever wonder why so many people expend so much effort
to have easy-to-use installers? It turns out that if installation is a
pain, many fewer people actually end up using your software. Gentoo is
more than just Portage.

Now, maybe people in Linux land have been too obsessed with having a
super-friendly installer - but it needs to be friendly-enough (and
compatible-enough) and it might be a good idea to take a fresh look at
how to streamline the Gentoo install experience.

Right now, installing Gentoo is a chore, and the many wonderful
choices of Gentoo end up making the install rather complicated. So I
definitely support ideas to help make our installation process
better/streamlined and less confusing. There are a lot of easy little
things that could be done.

-Daniel
--
gentoo-dev@gentoo.org mailing list

On 3/3/07, Daniel Robbins <drobbins.daniel@gmail.com> wrote:
> Right now, installing Gentoo is a chore, and the many wonderful
> choices of Gentoo end up making the install rather complicated. So I
> definitely support ideas to help make our installation process
> better/streamlined and less confusing. There are a lot of easy little
> things that could be done.

What do you think of a simplified handbook ? One that presents a lot
fewer choices to the user, in order to be less confusing. I don't mean
replacing the current handbook which is one great piece of work, but
writing a "Gentoo in 10 easy steps" kind of guide.

One of us may even have written one already. If not, I'm willing to
write or help writing one if that's considered a good idea.

Denis.
--
gentoo-dev@gentoo.org mailing list

On Sat, 2007-03-03 at 08:21 +0100, Denis Dupeyron wrote:
> What do you think of a simplified handbook ? One that presents a lot
> fewer choices to the user, in order to be less confusing. I don't mean
> replacing the current handbook which is one great piece of work, but
> writing a "Gentoo in 10 easy steps" kind of guide.
>
> One of us may even have written one already. If not, I'm willing to
> write or help writing one if that's considered a good idea.

http://www.gentoo.org/doc/en/gentoo-x86-quickinstall.xml

I'm not chiming in on the rest of the thread right now, since I am about
to head on a plane and won't be available for a bit, but to put it
simply, I have no plans on releasing *any* kind of nightly *anything* so
long as Release Engineering still gets minimal testing from only a
*tiny* subset of our developer pool when we are basically *begging* for
it. I'm not expecting the level of required testing to diminish just
because we do more builds, and in fact I expect it to increase. With
the current participation level, I just don't see it as possible.
Remember, we switched from quarterly to bi-annual releases for a reason.
We simply didn't have the man power, CPU power, nor time to do vigorous
enough testing in the much more shortened time frame. I'm going to be
asking for Release Testers again once I return, and if last year's turn
out was any indicator (50+ people volunteering, about 5 actually helping
*at all*), the chances of a project such as nightly builds ever taking
off is well beyond our means at this time.

--
Chris Gianelloni
Release Engineering Strategic Lead
Alpha/AMD64/x86 Architecture Teams
Games Developer/Council Member/Foundation Trustee
Gentoo Foundation

Denis Dupeyron wrote:
> On 3/3/07, Daniel Robbins <drobbins.daniel@gmail.com> wrote:
>> Right now, installing Gentoo is a chore, and the many wonderful
>> choices of Gentoo end up making the install rather complicated. So I
>> definitely support ideas to help make our installation process
>> better/streamlined and less confusing. There are a lot of easy little
>> things that could be done.
>
> What do you think of a simplified handbook ? One that presents a lot
> fewer choices to the user, in order to be less confusing. I don't mean
> replacing the current handbook which is one great piece of work, but
> writing a "Gentoo in 10 easy steps" kind of guide.
>
> One of us may even have written one already. If not, I'm willing to
> write or help writing one if that's considered a good idea.

Next time, read the documentation first.
http://www.gentoo.org/doc/en/list.xml

We've several quickstart/faq-type guides, and an alternate installation
howto. Man, I wish more developers would read the documentation, or at
least bother to get a general idea of what we have. There are some good
resources to be found with even minimal searching.

Anyway, re-writing the handbooks for multiple releases? Oh, hell no.
It's an incredible amount of work even with several GDP devs helping out
to get it ready for each point release, let alone when most of the team
is inactive, busy with real life, or otherwise unavailable.

Doing it nightly, weekly, or even monthly? That ain't gonna work. And
don't talk about "completely unsupported, so no need to write docs". The
main purpose here seems to be for hardware functionality, and that
changes from release to release. From trivial things like
s/dobladecenter/doslowusb to major things like new boot parameters and
RAID modules, to say nothing of how networking and baselayout change.
It's irresponsible to say to users, "Here, download this if you want to,
but you have to figure out how the heck it works, because we're not
telling, nyah nyah!" Users: "WTF no docs? Gentoo sux! YOU sux!"

So from the standpoint of the most active GDP member who'd have to write
the HBs 200 hours a week to keep up, it ain't gonna work too well. :)

On 3/3/07, Denis Dupeyron <calchan@gentoo.org> wrote:
> What do you think of a simplified handbook ? One that presents a lot
> fewer choices to the user, in order to be less confusing.

YES, it's needed. The handbook didn't turn out quite as I expected it
to. It should document a typical installation process with small links
to alternate approaches and options that a user might opt to follow.

And it should be one (web) page.

In my opinion.

-Daniel
--
gentoo-dev@gentoo.org mailing list

On Sat, 3 Mar 2007 01:34:36 -0700 "Daniel Robbins"
<drobbins.daniel@gmail.com> wrote:
> On 3/3/07, Denis Dupeyron <calchan@gentoo.org> wrote:
> > What do you think of a simplified handbook ? One that presents a lot
> > fewer choices to the user, in order to be less confusing.
>
> YES, it's needed. The handbook didn't turn out quite as I expected it
> to. It should document a typical installation process with small links
> to alternate approaches and options that a user might opt to follow.

I asked for this approach back when the handbook was first created. It
was rejected by the docs team for being "too complicated to maintain".
Following Sven's (I think...) suggestion, I instead ported the quick
install guide (which is one page, and doesn't go off on lots of
weird tangents) to the archs upon which I was working at the time.

--
Ciaran McCreesh
Mail : ciaranm at ciaranm.org
Web : http://ciaranm.org/
Paludis, the secure package manager : http://paludis.pioto.org/

On 3/3/07, Ciaran McCreesh <ciaranm@ciaranm.org> wrote:
> I asked for this approach back when the handbook was first created. It
> was rejected by the docs team for being "too complicated to maintain".
> Following Sven's (I think...) suggestion, I instead ported the quick
> install guide (which is one page, and doesn't go off on lots of
> weird tangents) to the archs upon which I was working at the time.

Yes, that was my request and I was told that this was the plan of
attack, but the end result looked nothing like this.

Just to be clear, I think the *official* documentation should be
simple, with a linear path and non-intrusive links for non-standard
stuff, and should fit on a single (Web) page.

I don't think that a quick install guide solves the problem. (I know
*you're* not saying it does, just pointing it out...) The official
documentation should be incredibly efficient to use, eliminating the
need for a quick install guide.

-Daniel
--
gentoo-dev@gentoo.org mailing list

Daniel Robbins wrote:
> Yes, that was my request and I was told that this was the plan of
> attack, but the end result looked nothing like this.
>
> Just to be clear, I think the *official* documentation should be
> simple, with a linear path and non-intrusive links for non-standard
> stuff, and should fit on a single (Web) page.
>
> I don't think that a quick install guide solves the problem. (I know
> *you're* not saying it does, just pointing it out...) The official
> documentation should be incredibly efficient to use, eliminating the
> need for a quick install guide.

I wholy agree on this point, yet I see a tiny problem. If you scrap all
the different paths in the handbook by keeping only one "standard" path,
I'd say about 50% of the whole thing is arch dependant :
- partitioning (i've never used anything other than x86/amd64)
- kernel setup
- bootloader

Common stuff:
- stage setup
- rsync, emerge system/world
- network setup
- additional utilities (such as cron, syslog)

Although I really would like to see a shorter guide, I think there are a
few issues (like these, but maybe others?) to layout first.

Ideas, comments ?

Rémi

PS, my 50% figure may not be very accurate :) don't feel offended if i'm
way off
--
gentoo-dev@gentoo.org mailing list

Daniel Robbins wrote:
> And it should be one (web) page.

http://www.gentoo.org/doc/en/handbook/handbook-x86.xml?full=1

--
Kind Regards,

Simon Stelling
Gentoo/AMD64
--
gentoo-dev@gentoo.org mailing list