[gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Ralph Seichter]

I know the Proxy Maintainer User Guide mentions mailing list patches as
a permissible alternative to GitHub pull requests. I also know that I
can locally discard these patches silently.

Still, I wonder why patches are posted here instead of using GitHub,
because there is no public review process going on. To me, email-based
patches on the Gentoo Proxy Maintainer mailing list are at best mildly
annoying, and at worst an attempt at jumping the queue, i.e. to get
one's changes processed faster than those made through GitHub. They
also, in my perception, have a smell of "ooh, look at me, I am working"
about them.

Thoughts?

-Ralph

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Joonas Niilola]

[-- Attachment #1.1: Type: text/plain, Size: 1141 bytes --]


On 6/7/20 2:45 AM, Ralph Seichter wrote:
> Still, I wonder why patches are posted here instead of using GitHub,
> because there is no public review process going on. To me, email-based
> patches on the Gentoo Proxy Maintainer mailing list are at best mildly
> annoying, and at worst an attempt at jumping the queue, i.e. to get
> one's changes processed faster than those made through GitHub. They
> also, in my perception, have a smell of "ooh, look at me, I am working"
> about them.

I also wonder that. Especially when I know some of the contributors DO
use Github. If you've read this list, you know my feelings about this
subject.

The automated CI check is a huge plus for both parties, and when I think
about it, the time spent should be rather equal whether you choose
git-send-email or git push. But the review and the 'fixing' part is
definitely cleaner with git push.

Now to your last sentence, I don't believe it's a matter of "attention
seeking", but more some disgust towards Github. But I'd also like hear
what's this all about. Would Gitlab be better than Github due to principles?

-- juippis



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 642 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Haelwenn (lanodan) Monnier]

[2020-06-07 10:01:43+0300] Joonas Niilola:
> On 6/7/20 2:45 AM, Ralph Seichter wrote:
> > Still, I wonder why patches are posted here instead of using GitHub,
> > because there is no public review process going on. To me, email-based
> > patches on the Gentoo Proxy Maintainer mailing list are at best mildly
> > annoying, and at worst an attempt at jumping the queue, i.e. to get
> > one's changes processed faster than those made through GitHub. They
> > also, in my perception, have a smell of "ooh, look at me, I am working"
> > about them.
> 
> I also wonder that. Especially when I know some of the contributors DO
> use Github. If you've read this list, you know my feelings about this
> subject.
> 
> The automated CI check is a huge plus for both parties, and when I think
> about it, the time spent should be rather equal whether you choose
> git-send-email or git push. But the review and the 'fixing' part is
> definitely cleaner with git push.
> 
> Now to your last sentence, I don't believe it's a matter of "attention
> seeking", but more some disgust towards Github. But I'd also like hear
> what's this all about. Would Gitlab be better than Github due to principles?
> 
> -- juippis

Since I'm the one which does send patches through email on proxy-maint:

- It's not to jump the queue, in fact I get about the same lag in my 
experience between GitHub PRs and using email. But it's much less predictable.

- I'd rather avoid GitHub completely, I do use it from time to time but because 
I have to.

And while sending via GitHub sometimes works I also got at least one PR later 
reverted because it didn't get a proper ACK from the maintainers, if I have to 
also ping them through email, why not just send the patch via email?
Specially as metadata.xml already gives me the emails.

There is also the problem of having to sync your repo to GitHub, 
if I forget to push gentoo's master to my fork I end up having to send all 
the past history manually, and I discovered this one quite later on.
This was a huge pain when I only had the choice between a slow connection or 
a metered one.
Also see https://forgeperf.org/ for how slow GitHub can get as just a web view.

Finally, I don't have this issue but github isn't available to everyone 
depending on where they live, and often a VPN isn't an option there, 
so another option needs to exists.

Best Regards,

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Ralph Seichter]

* Haelwenn Monnier:

> It's not to jump the queue, in fact I get about the same lag in my
> experience between GitHub PRs and using email. But it's much less
> predictable.

I don't understand the second sentence; what X is less predictable than
what Y, and why? As for the first sentence, I have my doubts, because
with less than 1% of contributors posting email patches, the ones that
do are very much "in your face" about it.

> I'd rather avoid GitHub completely, I do use it from time to time but
> because I have to.

I respect your right to dislike GitHub, but this is not about you alone
but about having automated CI checks and a unified means of letting the
proxy-maint members doing their volunteer job, in their own time, in a
comfortable way. Disclaimer: I do not speak for the P-M team, nor do I
claim to.

Another important aspect is that, by using GitHub, you allow only the
interested parties to subscribe to your pull requests (and I assume very
few would be interested). Posting patches here is, again, "in your
face", because rather than individually opting in on GitHub PRs, you are
forcing people to opt out every time you post a patch. I consider this
both impolite and annoying.

> And while sending via GitHub sometimes works I also got at least one
> PR later reverted because it didn't get a proper ACK from the
> maintainers, if I have to also ping them through email, why not just
> send the patch via email?

Because one of your PRs being met with a hiccup does not remotely
justify skipping the agreed-upon, preferred method of contribution
altogether.

There are OSS projects that use email-based patches successfully and
consistently, like "Notmuch" does. Howevery, Notmuch has the necessary
infrastructure to deal with these patches, and a culture of reviewing
these patches on the mailing list. Gentoo does not operate this way.

> There is also the problem of having to sync your repo to GitHub, if I
> forget to push gentoo's master to my fork I end up having to send all
> the past history manually, and I discovered this one quite later on.

You must be joking, right? Synchronising two Git remotes is no more
complicated than using "git send-email", and even if it was, learning to
use Git is mandatory. I remember nearly making a mess of my very first
pull request because of a Git-related mistake, which other people
obviously noticed. I assure you that I never made that mistake again,
and I am confident you can manage as well.

> Finally, I don't have this issue but github isn't available to
> everyone depending on where they live, and often a VPN isn't an
> option there, so another option needs to exists.

As I wrote right off the bat: I know patches *may* be posted here. That
does not imply they should, especially if the author, like yourself,
admittedly has the ability to use GitHub instead, which is clearly the
preferred method of contributing. If GitHub is an option, posting
patches here is both unnecessary and rude, in my opinion.

-Ralph

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Joonas Niilola]

[-- Attachment #1.1: Type: text/plain, Size: 3948 bytes --]


On 6/7/20 8:24 PM, Ralph Seichter wrote:
> * Haelwenn Monnier:
>
>> It's not to jump the queue, in fact I get about the same lag in my
>> experience between GitHub PRs and using email. But it's much less
>> predictable.
> I don't understand the second sentence; what X is less predictable than
> what Y, and why? As for the first sentence, I have my doubts, because
> with less than 1% of contributors posting email patches, the ones that
> do are very much "in your face" about it.
>
>> I'd rather avoid GitHub completely, I do use it from time to time but
>> because I have to.
> I respect your right to dislike GitHub, but this is not about you alone
> but about having automated CI checks and a unified means of letting the
> proxy-maint members doing their volunteer job, in their own time, in a
> comfortable way. Disclaimer: I do not speak for the P-M team, nor do I
> claim to.
>
> Another important aspect is that, by using GitHub, you allow only the
> interested parties to subscribe to your pull requests (and I assume very
> few would be interested). Posting patches here is, again, "in your
> face", because rather than individually opting in on GitHub PRs, you are
> forcing people to opt out every time you post a patch. I consider this
> both impolite and annoying.
>
>> And while sending via GitHub sometimes works I also got at least one
>> PR later reverted because it didn't get a proper ACK from the
>> maintainers, if I have to also ping them through email, why not just
>> send the patch via email?
> Because one of your PRs being met with a hiccup does not remotely
> justify skipping the agreed-upon, preferred method of contribution
> altogether.
>
> There are OSS projects that use email-based patches successfully and
> consistently, like "Notmuch" does. Howevery, Notmuch has the necessary
> infrastructure to deal with these patches, and a culture of reviewing
> these patches on the mailing list. Gentoo does not operate this way.
>
>> There is also the problem of having to sync your repo to GitHub, if I
>> forget to push gentoo's master to my fork I end up having to send all
>> the past history manually, and I discovered this one quite later on.
> You must be joking, right? Synchronising two Git remotes is no more
> complicated than using "git send-email", and even if it was, learning to
> use Git is mandatory. I remember nearly making a mess of my very first
> pull request because of a Git-related mistake, which other people
> obviously noticed. I assure you that I never made that mistake again,
> and I am confident you can manage as well.
>
>> Finally, I don't have this issue but github isn't available to
>> everyone depending on where they live, and often a VPN isn't an
>> option there, so another option needs to exists.
> As I wrote right off the bat: I know patches *may* be posted here. That
> does not imply they should, especially if the author, like yourself,
> admittedly has the ability to use GitHub instead, which is clearly the
> preferred method of contributing. If GitHub is an option, posting
> patches here is both unnecessary and rude, in my opinion.
>
> -Ralph

Hey,

there are some bits in this message where the tone comes out a bit
hostile. I'd hope we can continue this discussion in a civil manner.

There were some interesting points already and I can see the bandwith be
a problem for some.

In the end, we are *all* trying to make Gentoo better, right? I am
thankful of every contribution, and this mailing list is low traffic at
the moment. If Github is not an option, please don't let that stop you
from contributing. You can of course send git-format patches directly to
proxy-maint@gentoo.org, too.

But I'm still interested about the discussion _why_ Github doesn't work
for everyone. We can maybe find some new ideas or new ways, when we are
aware of all the problems.

-- juippis



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 642 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Robert Günzler]

On Sun Jun 7, 2020 at 11:15 PM CEST, Joonas Niilola wrote:
> In the end, we are *all* trying to make Gentoo better, right? I am
> thankful of every contribution, and this mailing list is low traffic at
> the moment. If Github is not an option, please don't let that stop you
> from contributing. You can of course send git-format patches directly to
> proxy-maint@gentoo.org, too.

This 100%! To me that also includes not making the maintainers life
harder though...

> But I'm still interested about the discussion _why_ Github doesn't work
> for everyone. We can maybe find some new ideas or new ways, when we are
> aware of all the problems.

I honestly think it doesn't matter so much _why_ people prefer something else
over GH. How much are you considering to switch to yet another
solution after all... 
And I bet even that wouldn't make everybody happy :P

I'm fairly new to Gentoo, so I'm actually confused that https://packages.gentoo.org 
always links to https://cgit.gentoo.org/repo/gentoo.git - I thought that
is the "source-of-truth" place then... but patches are going into the GH
repo. Who is actually sycing all that back to cgit then?

Maybe a solution would be to make the review + dev workflows
independent of the git solution?

In response to me asking about https://sourcehut.org, in a recent thread 
you described your workflow:

On Sat Jun 6, 2020 at 12:57 PM CEST, Joonas Niilola wrote:
> 1: Open https://github.com/gentoo/gentoo/pull/16029
>
> 2: Launch a container with up-to-date ~ stage3,
>
> 3: "fullget.sh 16029" (a bash script to download
> https://patch-diff.githubusercontent.com/raw/gentoo/gentoo/pull/XXXXXX.pa=
> tch
> and apply it to containers portage tree)(fullget.sh, because there is
> another script to cherry-pick commits),
>
> 4: pkgcheck scan --commits,
>
> 5: depmerge.sh package (emerge -1av --onlydeps --with-test-deps "$@"),
>
> 6: testmerge.sh package (emerge -1av with
> https://wiki.gentoo.org/wiki/Package_testing#make.conf these settings),
>
> 7: If all is good, switch to desktop shell and ::gentoo repo, use "pram
> 16029" (https://packages.gentoo.org/packages/app-portage/pram)
>
> (I'm looking to automate steps 3-6 with one command)
>
>
> With e-mail patches, I have to copypaste them manually into container,
> and then apply them manually to ::gentoo repo one-by-one. It's not a lot
> extra work, but it slows down the process a bit. And reviewing
> contribution is a bit tedious compared to Github, especially if there is
> a big contribution.

How much value would there be for you, as a first iteration of a
solution, to get patches on the mailing-list automatically submitted as
a PR (done by a bot) on GH?

	robert

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Randy Barlow]

There is a contingent of kernel developers in the Fedora space who 
prefer to use e-mail for patches, so this project was created to bridge 
e-mailed patches with GitLab pull requests:

https://github.com/jeremycline/patchlab

I've never used it and don't know a ton about it, but AIUI it can do a 
bi-directional sync between e-mailed patches and GitLab pull requests. I 
happen to be friends with the author, and he told me once that he thinks 
it'd be pretty easy to make it work with GitHub as well (i.e., most of 
the code is dealing with mapping e-mails to the concept of pull 
requests, not the GitLab API per se). So perhaps that could be adapted 
to suit more people's desires here?

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Jimi Huotari]

[-- Attachment #1: Type: text/plain, Size: 590 bytes --]

On Mon, 08 Jun 2020 10:42:23 +0200
Robert Günzler <robert@gnzler.io> wrote:

> I'm fairly new to Gentoo, so I'm actually confused that
> https://packages.gentoo.org always links to
> https://cgit.gentoo.org/repo/gentoo.git - I thought that is the
> "source-of-truth" place then... but patches are going into the GH repo.
> Who is actually sycing all that back to cgit then?

Welcome to Gentoo!

As far as I'm aware of things and stuff, GitHub is only used as a mirror
and for code review.

Nothing is actually merged via/using GitHub itself into the official
repositories.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 981 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Joonas Niilola]

[-- Attachment #1.1: Type: text/plain, Size: 1838 bytes --]


On 6/8/20 11:42 AM, Robert Günzler wrote:
> I honestly think it doesn't matter so much _why_ people prefer something else
> over GH. How much are you considering to switch to yet another
> solution after all... 
> And I bet even that wouldn't make everybody happy :P

I wouldn't mind switching to Gentoo-hosted alternative, that does the
same. But I imagine it'd require too much work, and why replicate
something that already exists and works well.


>
> I'm fairly new to Gentoo, so I'm actually confused that https://packages.gentoo.org 
> always links to https://cgit.gentoo.org/repo/gentoo.git - I thought that
> is the "source-of-truth" place then... but patches are going into the GH
> repo. Who is actually sycing all that back to cgit then?
>
As chiitoo said, Github is just used to mirror the tree and enable
creating pull requests. Github also allows easy-to-follow review
process, and CI checks before the patch is merged. Basically you take
https://patch-diff.githubusercontent.com/raw/gentoo/gentoo/pull/16110.patch
and apply it to real git tree.


>
> In response to me asking about https://sourcehut.org, in a recent thread 
> you described your workflow:
>
>
> How much value would there be for you, as a first iteration of a
> solution, to get patches on the mailing-list automatically submitted as
> a PR (done by a bot) on GH?
>
> 	robert
If the person posting it doesn't use Github, then reviewing the pull
request goes to vain. Unless it somehow returns the feedback to sender,
but that could get dirty real fast.  We sometimes give a lot of feedback
and require fixes before it can be merged. Other than that, it's an
interesting idea, but it adds more hoops to the process and that adds
probability to have mistakes / lose the contribution on the way.

-- juippis



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 642 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[tastytea]

On 2020-06-07 10:01+0300 Joonas Niilola <juippis@gentoo.org> wrote:

> […]
>
> Now to your last sentence, I don't believe it's a matter of "attention
> seeking", but more some disgust towards Github. But I'd also like hear
> what's this all about. Would Gitlab be better than Github due to
> principles?

Some of the problems with GitHub (and its parent company Microsoft)
are:

* They are doing business with ICE¹, an organisation which runs
  concentrations camps² and whose director used the Nuremberg defense³
  when asked about it. Here is a list of articles on the subject: ⁴.

* Microsoft collaborated with the NSA on the illegal “PRISM” program⁵
  and collaborates with the U.S. military to conduct mass murder.

* GitHub is blocking developers in countries facing US trade sanctions⁶.

* GitHub has been called out repeatedly⁷ for sexism⁸.

And no, GitLab (the company) would not be much better⁹.

¹ U.S. Immigration and Customs Enforcement
² <https://www.esquire.com/news-politics/a27813648/concentration-camps-southern-border-migrant-detention-facilities-trump/>
³ <https://mavenroundtable.io/theintellectualist/news/ex-ice-director-ice-can-t-be-equated-to-nazis-since-we-re-just-following-orders-gUCVH4E1CUi5Y73kZqpeNA>
⁴ <https://sneak.berlin/20200307/the-case-against-microsoft-and-github/#collaboration-with-ice-who-runs-concentration-camps>
⁵ <https://en.wikipedia.org/wiki/PRISM_(surveillance_program)>
⁶ <https://www.zdnet.com/article/github-starts-blocking-developers-in-countries-facing-us-trade-sanctions/>
⁷ <https://www.cnet.com/news/github-were-sorry-for-secrecy-with-sexism-claims-inquiry/>
⁸ <https://where.coraline.codes/blog/my-year-at-github/>
⁹ <https://www.theregister.com/2020/02/06/gitlab_sales_women/>

-- 
Get my PGP key with `gpg --locate-keys tastytea@tastytea.de` or at
<https://tastytea.de/tastytea.asc>.

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Joonas Niilola]

[-- Attachment #1.1: Type: text/plain, Size: 1856 bytes --]


On 6/7/20 7:58 PM, tastytea wrote:
>
> Some of the problems with GitHub (and its parent company Microsoft)
> are:
>
> * They are doing business with ICE¹, an organisation which runs
>   concentrations camps² and whose director used the Nuremberg defense³
>   when asked about it. Here is a list of articles on the subject: ⁴.
>
> * Microsoft collaborated with the NSA on the illegal “PRISM” program⁵
>   and collaborates with the U.S. military to conduct mass murder.
>
> * GitHub is blocking developers in countries facing US trade sanctions⁶.
>
> * GitHub has been called out repeatedly⁷ for sexism⁸.
>
> And no, GitLab (the company) would not be much better⁹.
>
> ¹ U.S. Immigration and Customs Enforcement
> ² <https://www.esquire.com/news-politics/a27813648/concentration-camps-southern-border-migrant-detention-facilities-trump/>
> ³ <https://mavenroundtable.io/theintellectualist/news/ex-ice-director-ice-can-t-be-equated-to-nazis-since-we-re-just-following-orders-gUCVH4E1CUi5Y73kZqpeNA>
> ⁴ <https://sneak.berlin/20200307/the-case-against-microsoft-and-github/#collaboration-with-ice-who-runs-concentration-camps>
> ⁵ <https://en.wikipedia.org/wiki/PRISM_(surveillance_program)>
> ⁶ <https://www.zdnet.com/article/github-starts-blocking-developers-in-countries-facing-us-trade-sanctions/>
> ⁷ <https://www.cnet.com/news/github-were-sorry-for-secrecy-with-sexism-claims-inquiry/>
> ⁸ <https://where.coraline.codes/blog/my-year-at-github/>
> ⁹ <https://www.theregister.com/2020/02/06/gitlab_sales_women/>
>
I believe similar accusations can be found for every corporation. If you
dig deep enough, doubt there are truly clean slates for any major IT
actor. Judging by the links, at least Github seems to drive for better
equality.
"GitHub is a better place now."

-- juippis



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 642 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[tastytea]

On 2020-06-09 09:10+0300 Joonas Niilola <juippis@gentoo.org> wrote:

> […]
> I believe similar accusations can be found for every corporation. If
> you dig deep enough, doubt there are truly clean slates for any major
> IT actor. Judging by the links, at least Github seems to drive for
> better equality.
> "GitHub is a better place now."

Every other corporation is happily taking money from fascist
organizations? I doubt that. 
The obvious and most sustainable solution is to self-host, which is
being discussed¹ on gentoo-dev right now by the way.

Kind regards, tastytea

¹ <https://archives.gentoo.org/gentoo-dev/message/516962fbfea5d60dd24ded8b12958584>

-- 
Get my PGP key with `gpg --locate-keys tastytea@tastytea.de` or at
<https://tastytea.de/tastytea.asc>.

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Michał Górny]

[-- Attachment #1: Type: text/plain, Size: 1225 bytes --]

On Tue, 2020-06-09 at 13:37 +0200, tastytea wrote:
> On 2020-06-09 09:10+0300 Joonas Niilola <juippis@gentoo.org> wrote:
> 
> > […]
> > I believe similar accusations can be found for every corporation. If
> > you dig deep enough, doubt there are truly clean slates for any major
> > IT actor. Judging by the links, at least Github seems to drive for
> > better equality.
> > "GitHub is a better place now."
> 
> Every other corporation is happily taking money from fascist
> organizations? I doubt that. 
> The obvious and most sustainable solution is to self-host, which is
> being discussed¹ on gentoo-dev right now by the way.
> 

There's a long way from 'being discussed' and a working solution.  Infra
hasn't managed to get a working solution for years, and now instead of
deploying GitLab like every other successful projects they're trying
hard to prove that some random half-maintained package will do better.

Please don't forget that even if we get a good solution deployed, it
will require someone to do all the work to port our scripts from GitHub
API to whatever deployed solution's API.  And that's only assuming it
will have good enough API.

-- 
Best regards,
Michał Górny


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 618 bytes --]

Re: [gentoo-proxy-maint] Posting patches on this mailing list instead of using GitHub

[Ralph Seichter]

* Joonas Niilola:

> I don't believe it's a matter of "attention seeking", but more some
> disgust towards Github.

Possible, but I am not so sure. ;-)

> But I'd also like hear what's this all about. Would Gitlab be better
> than Github due to principles?

I just checked my projects, and found Git remotes for bitbucket.org,
gitlab.com, github.com, a couple of self-hosted GitLab instances
(including my company's own). From a contributor's perspective, they
differ very little. The collaboration aspects are also similar enough.
I should mention that there are also Gitolite-based remotes, but the
ones I use are pure storage and don't offer any collaboration tools.

I find the main differences in CI/CD, and that's where organisations
seem to make their choices when they are not self-hosting.

-Ralph