[gentoo-dev] Optimizing toe stepping

[gentoo-dev] Optimizing toe stepping

[Jason A. Donenfeld]

Hey guys,

Every other day on IRC, I see people arguing about touching each
others packages, despite our policies against it. (Sometimes it's even
me who's doing the touching!) My instinctive reaction is always,
"can't everybody calm down and be happy somebody is doing your work
for you?" The answer to this question is, of course, that it depends
who the developer is and how competent you are.

So, nothing new here. I don't want to bikeshed about it. Business as usual.

Perhaps we can come up with something more formal for solving this,
that works a bit better than my package-policy.txt idea (see other
thread).

A few have proposed in the past adding this kind of "attitude
information" to metadata.xml. Does anybody have any concrete proposals
for what this would look like?

Thanks,
Jason

-- 
Jason A. Donenfeld
Gentoo Linux Security & Infrastructure
zx2c4@gentoo.org
www.zx2c4.com
zx2c4.com/keys/A28BEDE08F1744E16037514806C4536755758000.asc

Re: [gentoo-dev] Optimizing toe stepping

[Nick Vinson]

[-- Attachment #1.1: Type: text/plain, Size: 1669 bytes --]

On 11/03/2016 03:16 AM, Jason A. Donenfeld wrote:
> Hey guys,
> 
> Every other day on IRC, I see people arguing about touching each
> others packages, despite our policies against it. (Sometimes it's even
> me who's doing the touching!) My instinctive reaction is always,
> "can't everybody calm down and be happy somebody is doing your work
> for you?" The answer to this question is, of course, that it depends
> who the developer is and how competent you are.
> 
> So, nothing new here. I don't want to bikeshed about it. Business as usual.
> 
> Perhaps we can come up with something more formal for solving this,
> that works a bit better than my package-policy.txt idea (see other
> thread).
> 
> A few have proposed in the past adding this kind of "attitude
> information" to metadata.xml. Does anybody have any concrete proposals
> for what this would look like?

I must have missed the latest IRC conversation on this.  Was the issue
really about someone else doing the work or was it about someone
committing changes without informing the actual maintainer(s) first?

If it's the first, there's no issue and nothing needs to be done in my
opinion (well maybe the maintainer(s) should take note and solve the
issue(s) faster next time).  If it's the latter, then I think a much
easier solution would be to file a PR, write a bug, or send an email
before committing the change.  None of those options are particularly
difficult to do and it keeps the maintainer in the loop.

Just doing that one little thing would have prevented or shutdown the
arguments I have seen.

Thanks,
Nicholas Vinson

> 
> Thanks,
> Jason
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [gentoo-dev] Optimizing toe stepping

[Jason A. Donenfeld]

On Thu, Nov 3, 2016 at 1:15 PM, Nick Vinson <nvinson234@gmail.com> wrote:
> Just doing that one little thing would have prevented or shutdown the
> arguments I have seen.

Yes, obviously of course.

But sometimes it's just easier and quicker to meddle in somebody
else's ebuild to improve a particular situation. The question is thus
how do we make this permissible? My proposal is adding a metadata.xml
tag -- <yes-you-can-meddle-in-my-affairs>.

Jason

Re: [gentoo-dev] Optimizing toe stepping

[Ian Stakenvicius]

> On Nov 3, 2016, at 8:21 AM, Jason A. Donenfeld <zx2c4@gentoo.org> wrote:
> 
>> On Thu, Nov 3, 2016 at 1:15 PM, Nick Vinson <nvinson234@gmail.com> wrote:
>> Just doing that one little thing would have prevented or shutdown the
>> arguments I have seen.
> 
> Yes, obviously of course.
> 
> But sometimes it's just easier and quicker to meddle in somebody
> else's ebuild to improve a particular situation. The question is thus
> how do we make this permissible? My proposal is adding a metadata.xml
> tag -- <yes-you-can-meddle-in-my-affairs>.
> 
> Jason
> 

Although metadata.xml is one way to do this, since it is more of a social thing than a technical one I think it might be better to wikify it instead -- each dev can list their "please fix my package" preferences in a per package or per anything-with-them-as-maintainer spec in one location.

The previous round of this had a bunch of devs make comments to effect of its fine for any dev to commit changes as long as the committee takes responsibility for it, it'll take a bit of time but we can probably compile those from irc logs and -dev@ and start the list.  Thoughts?

Re: [gentoo-dev] Optimizing toe stepping

[Rich Freeman]

On Thu, Nov 3, 2016 at 9:45 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
>
>
> Although metadata.xml is one way to do this, since it is more of a social thing than a technical one I think it might be better to wikify it instead -- each dev can list their "please fix my package" preferences in a per package or per anything-with-them-as-maintainer spec in one location.
>

I tend to think that metadata is the right place for a couple of reasons:

1.  Somebody who discovers an ebuild with an issue/etc is probably
sitting right in the directory with the metadata file, so the
information is readily at hand.
2.  If somebody was going to have to reach out to the maintainer, the
metadata file would tell them who the maintainer is (both in terms of
projects and individuals).
3.  The file could potentially contain package-specific maintenance
information.  Sure, you can stick a page on a wiki that says "for
rich0 in general feel free to touch anything, but be aware that mythtv
upstream is picky about xyz, and be aware that the android sdk has
issue xyz, ..."  For somebody with their fingers on a lot of packages
you could end up either writing a book, or just leaving it all out
which could result in people making the same mistakes over and over,
or devs might just opt out of having others touch their stuff because
it is too much of a PITA to explain it all.  With the metadata
approach you only define package-level detail.  So, if one package is
hands-off, then you simply state so or fail to give permission to
touch it.  You could provide other background that is relevant to the
specific package.

I think the main issue is where to put it in the schema.  For the
proponents I'd suggest just putting a stick in the mud and see if
anybody complains.

-- 
Rich

Re: [gentoo-dev] Optimizing toe stepping

[William Hubbs]

[-- Attachment #1: Type: text/plain, Size: 1989 bytes --]

On Thu, Nov 03, 2016 at 09:54:39AM -0400, Rich Freeman wrote:
> On Thu, Nov 3, 2016 at 9:45 AM, Ian Stakenvicius <axs@gentoo.org> wrote:
> >
> >
> > Although metadata.xml is one way to do this, since it is more of a social thing than a technical one I think it might be better to wikify it instead -- each dev can list their "please fix my package" preferences in a per package or per anything-with-them-as-maintainer spec in one location.
> >
> 
> I tend to think that metadata is the right place for a couple of reasons:
> 
> 1.  Somebody who discovers an ebuild with an issue/etc is probably
> sitting right in the directory with the metadata file, so the
> information is readily at hand.
> 2.  If somebody was going to have to reach out to the maintainer, the
> metadata file would tell them who the maintainer is (both in terms of
> projects and individuals).
> 3.  The file could potentially contain package-specific maintenance
> information.  Sure, you can stick a page on a wiki that says "for
> rich0 in general feel free to touch anything, but be aware that mythtv
> upstream is picky about xyz, and be aware that the android sdk has
> issue xyz, ..."  For somebody with their fingers on a lot of packages
> you could end up either writing a book, or just leaving it all out
> which could result in people making the same mistakes over and over,
> or devs might just opt out of having others touch their stuff because
> it is too much of a PITA to explain it all.  With the metadata
> approach you only define package-level detail.  So, if one package is
> hands-off, then you simply state so or fail to give permission to
> touch it.  You could provide other background that is relevant to the
> specific package.

I'm adding Robin to this thread, because he wrote up a similar proposal
a while back. I don't remember what happened to it at the time, and I do
not have a link to it.

I am also in favor of the metadata approach.

William


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 163 bytes --]

Re: [gentoo-dev] Optimizing toe stepping

[William L. Thomson Jr.]

[-- Attachment #1: Type: text/plain, Size: 846 bytes --]

On Thursday, November 3, 2016 9:14:56 AM EDT William Hubbs wrote:
> 
> I am also in favor of the metadata approach.

The only downfall is you end up with lots of package/developer specific 
policies which may become complex and a nuisance. It is likely better to have 
more of a global policy.

Also it may start inching things towards restricting areas of the tree to said 
members of teams etc. Which starts to go in the opposite direction of Gentoo's 
roots of wide open access. To specifically encourage others to touch others 
stuff, and help everything move along faster. Development wise, not ricing out 
binaries...

I do think something like a maintainers notes URI in a metadata.xml file would 
be of use for many reasons. Not putting a bunch of rules or stuff in the 
metadata.xml file.

-- 
William L. Thomson Jr.

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 163 bytes --]

Re: [gentoo-dev] Optimizing toe stepping

[Robin H. Johnson]

On Thu, Nov 03, 2016 at 05:36:03PM -0400, William L. Thomson Jr. wrote:
> On Thursday, November 3, 2016 9:14:56 AM EDT William Hubbs wrote:
I have posted about this subject before, and I'll go and dust off the
proposal with the last changes added, when I have time soon hopefully.

> > I am also in favor of the metadata approach.
> The only downfall is you end up with lots of package/developer specific 
> policies which may become complex and a nuisance. It is likely better to have 
> more of a global policy.
Sure, global policies help (as a sane default), but being able to tell
people the most common cases is important:
1. Yes, touch it, but please ask me (DEFAULT, including timeout value)
2. Yes, go ahead and touch this package, and don't ask me
3. danger will robinson! this is fragile!

I'm going to set #2 on the great majority of the 200+ packages that I
directly maintain.

Make it easy to open a package up to MORE changes, and hard to restrict.

-- 
Robin Hugh Johnson
Gentoo Linux: Dev, Infra Lead, Foundation Trustee & Treasurer
E-Mail   : robbat2@gentoo.org
GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85
GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136

Re: [gentoo-dev] Optimizing toe stepping

[Austin English]

[-- Attachment #1.1: Type: text/plain, Size: 1418 bytes --]

On 11/03/2016 04:49 PM, Robin H. Johnson wrote:
> On Thu, Nov 03, 2016 at 05:36:03PM -0400, William L. Thomson Jr. wrote:
>> On Thursday, November 3, 2016 9:14:56 AM EDT William Hubbs wrote:
> I have posted about this subject before, and I'll go and dust off the
> proposal with the last changes added, when I have time soon hopefully.
> 
>>> I am also in favor of the metadata approach.
>> The only downfall is you end up with lots of package/developer specific 
>> policies which may become complex and a nuisance. It is likely better to have 
>> more of a global policy.
> Sure, global policies help (as a sane default), but being able to tell
> people the most common cases is important:
> 1. Yes, touch it, but please ask me (DEFAULT, including timeout value)
> 2. Yes, go ahead and touch this package, and don't ask me
> 3. danger will robinson! this is fragile!
> 
> I'm going to set #2 on the great majority of the 200+ packages that I
> directly maintain.
> 
> Make it easy to open a package up to MORE changes, and hard to restrict.
> 

+1

Doing a lot of tree wide cleanups, it is a pain to have to track down
maintainers to ACK simple changes at scale (e.g.,
runscript->openrc-run). If a portion of the packages had a notation that
I could poke without pinging, it would speed that up quite a bit.

-- 
-Austin

Austin English
Gentoo Developer
GPG: 00B3 2957 B94B F3E1


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [gentoo-dev] Optimizing toe stepping

[Andrew Savchenko]

[-- Attachment #1: Type: text/plain, Size: 1482 bytes --]

On Thu, 3 Nov 2016 21:49:16 +0000 Robin H. Johnson wrote:
> On Thu, Nov 03, 2016 at 05:36:03PM -0400, William L. Thomson Jr. wrote:
> > On Thursday, November 3, 2016 9:14:56 AM EDT William Hubbs wrote:
> I have posted about this subject before, and I'll go and dust off the
> proposal with the last changes added, when I have time soon hopefully.
> 
> > > I am also in favor of the metadata approach.
> > The only downfall is you end up with lots of package/developer specific 
> > policies which may become complex and a nuisance. It is likely better to have 
> > more of a global policy.
> Sure, global policies help (as a sane default), but being able to tell
> people the most common cases is important:
> 1. Yes, touch it, but please ask me (DEFAULT, including timeout value)
> 2. Yes, go ahead and touch this package, and don't ask me
> 3. danger will robinson! this is fragile!
> 
> I'm going to set #2 on the great majority of the 200+ packages that I
> directly maintain.
> 
> Make it easy to open a package up to MORE changes, and hard to restrict.
 
Looks like a reasonable approach. Probably it is worth to set
global policy describing these options and setting default
value. Should this issue be discussed with the council now?

As for implementation, repoman checks accepting this field should
be added, and skel.metadata.xml may be updated. I see no other
required changes, though I may miss something.

Best regards,
Andrew Savchenko

[-- Attachment #2: Type: application/pgp-signature, Size: 819 bytes --]

Re: [gentoo-dev] Optimizing toe stepping

[Rich Freeman]

On Thu, Nov 3, 2016 at 5:36 PM, William L. Thomson Jr.
<wlt-ml@o-sinc.com> wrote:
> On Thursday, November 3, 2016 9:14:56 AM EDT William Hubbs wrote:
>>
>> I am also in favor of the metadata approach.
>
> Also it may start inching things towards restricting areas of the tree to said
> members of teams etc. Which starts to go in the opposite direction of Gentoo's
> roots of wide open access.

Actually, the default right now is somewhat restrictive.  I advocate
this mainly because it will probably tend to open things up.  By
default today you aren't supposed to just go touching things without
pinging somebody, and that is probably a reasonable default, but
having an easy way to open things up where it makes sense is a good
idea.

-- 
Rich