Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

time to tag/release 2.20? #185

Open
zenczykowski opened this issue Aug 15, 2022 · 46 comments
Open

time to tag/release 2.20? #185

zenczykowski opened this issue Aug 15, 2022 · 46 comments

Comments

@zenczykowski
Copy link

in particular I'm interested in getting pref64 into fedora...

@agowa
Copy link

agowa commented Aug 30, 2022

Same. Also I'd like AdvCaptivePortalAPI to get into distros.

@Neustradamus
Copy link
Member

@robbat2, @stappersg: I think it is needed before Debian 12 freeze.

@aalmenar
Copy link

aalmenar commented Nov 1, 2022

I believe a new release is needed since the last one was more than two years ago. Pref64 being one of the lastest additions and some bug fixes.

@kloczek
Copy link

kloczek commented Dec 29, 2022

+1 😋

@alexhaydock
Copy link

Is there a roadmap to be satisfied before the tagging of a new release?

Seems like there's been a lot of development in the main branch since the last release 2 years ago. Or is it the expectation that downstream distributions should start building from git rather than tagged releases at this point?

@wuch-g2v
Copy link

+1

@manonfgoo
Copy link

manonfgoo commented Sep 3, 2023

Yes, I'd like to get RemoveAdvOnExit into freebsd ports

@oskar456
Copy link

oskar456 commented Nov 5, 2023

Ping. Is there anything we could help with getting a released version with all new features?

@Neustradamus
Copy link
Member

@robbat2: More and more people request a new build, the latest current build is 2.19 (2020-09-20).
I think it is good to do a 2.20, can you do it?

The current changes, 114 commits:

Contributors since 2.19 (in more you) :)
@nomis, @stappersg, @robbat2, @RICCIARDI-Adrien, @haegar, @samboyles1, @steglicd, @jpds, @andrew-sayers, @juhamatk, @manonfgoo, @gaoxingwang, @yas-nyan, @zajdee, @initramfs, @kepstin, @prometheanfire, @thesamesam, @mpontillo, @jetomit

Thanks in advance.

cc @reubenhwk (old maintainer)

@reubenhwk
Copy link
Collaborator

reubenhwk commented Nov 12, 2023 via email

@reubenhwk
Copy link
Collaborator

reubenhwk commented Nov 12, 2023 via email

@reubenhwk
Copy link
Collaborator

reubenhwk commented Nov 12, 2023 via email

@stappersg
Copy link
Member

stappersg commented Nov 12, 2023 via email

@manonfgoo
Copy link

manonfgoo commented Nov 12, 2023 via email

@robbat2
Copy link
Member

robbat2 commented Nov 12, 2023

I'll tag a release candidate version after #210 is approved & merged.

I heard a non-public report that newer kernels might have an interaction, so I want to force more testing rather than saying "this is a good public release" - plus we have outstanding issues where we have asked users to test, and they never got back to us.

@Neustradamus
Copy link
Member

@robbat2 has done a 2.20 RC1 release tag:

Can you try and please do a feedback?

@alexhaydock
Copy link

@robbat2 has done a 2.20 RC1 release tag:

Can you try and please do a feedback?

I've flagged this as out-of-date in Alpine Linux edge in the hope it can be brought into edge, or into the testing repo so I can do some testing in my production environment that's based around Alpine Edge.

If not, I'll try and build this manually at some point soon and do some testing, though I'd need to find more time for that.

@kloczek
Copy link

kloczek commented Nov 17, 2023

@robbat2 has done a 2.20 RC1 release tag:

Please change versioning convention.
Instead v2.20_rc1 better would be use just v2.20.rc1.
More about that versioning you can find on https://discourse.gnome.org/t/new-gnome-versioning-scheme/4235.

@robbat2
Copy link
Member

robbat2 commented Nov 20, 2023

@kloczek why has GNOME moved away from decades of using -/_ for the rcX slugs?

Gentoo for example codified common upstream practices to have the underscore separating the core version number.
https://dev.gentoo.org/~ulm/pms/head/pms.html#section-3.2

@robbat2
Copy link
Member

robbat2 commented Nov 20, 2023

Separately, we need to improve the automation to upload proper dist tarballs, since the GitHub release system automatically used the tip of the tree, rather than the files I wanted to upload.

@stappersg
Copy link
Member

@robbat2 has done a 2.20 RC1 release tag:

Please change versioning convention.
Instead v2.20_rc1 better would be use just v2.20.rc1.

FWIW, I'm fine with v2.20.n as release candidates and v2.20.n+1 as actual release. So more release early, release often.

Less subtle: Don't bother about release candidates.

@kloczek
Copy link

kloczek commented Nov 20, 2023

@kloczek why has GNOME moved away from decades of using -/_ for the rcX slugs?

Because:

  • - is character forbidden by many package manager software to use in version string
  • rc*, alpha*, beta*, 0 strings are in alphabetical order

Both reasons are to make versioning packaging friendly.

@kloczek
Copy link

kloczek commented Nov 20, 2023

FWIW, I'm fine with v2.20.n as release candidates and v2.20.n+1 as actual release. So more release early, release often.

That is why better is use sequence rc*, alpha*, beta*, 0 ... <n> or subset of that sequence.

@alexhaydock
Copy link

My initial testing suggests this builds and is stable on Alpine Linux edge, and I've been able to test and validate that PREF64 extensions are working, as a particular feature which was added between 2.19 and 2.20_RC1

image

I'm pleased to note that this also builds without needing fix-alpine-plz.patch which Alpine was shipping as a backport of a fix (06689f8) which was added after 2.19 was tagged.

I'll continue to run this for a while and report back if anything turns up.

@alexhaydock
Copy link

Where are we at on testing for this one? Is there anything to be satisfied before bringing 2.20 out of RC status?

I notice #227 and #222 are open and were opened after the RC was tagged, but #227 is explicitly marked as needing to wait until after the current release, and I can't see #222 as a showstopper that would block release.

@cirkku
Copy link

cirkku commented Jan 28, 2024

Any luck in getting this released? My impression is that its stable and getting an official release would be necessary for integrating RFC 8781 into VyOS nightly builds.

@heliosfa
Copy link

heliosfa commented Feb 25, 2024

I've been doing some testing of this as I am quite keen on the PREF64 support and it all seems happy under Linux (Debian 12), but FreeBSD 14 still needs the patches discussed in #145 from their ports repo or RADVD doesn't consistently respond to RA solicitations. Not much is needed and it all seems happy on FreeBSD 14.

@Neustradamus
Copy link
Member

@robbat2: It is a good time to do it?

cc: @reubenhwk.

@paravoid
Copy link

Thank you for your work on radvd! We're now past the one year anniversary of 2.20-rc1. Any news here? Any way folks can help?

@Neustradamus
Copy link
Member

@Neustradamus
Copy link
Member

@robbat2, @stappersg, @reubenhwk: It important to have improvements before Debian 13 freeze.

@robbat2
Copy link
Member

robbat2 commented Nov 27, 2024

@Neustradamus i'm not seeing any announced freeze from Debian Release Management? https://release.debian.org/
Last update there was 2023/12/16

@Neustradamus
Copy link
Member

Like every time, it will begin in January, it will be in January 2025, in few weeks:

@melroy89
Copy link
Contributor

melroy89 commented Dec 5, 2024

Latest release tag is from 2021, so I'm ready for a new release :)

@stappersg
Copy link
Member

Help with "github.com actions", #236, is appreciated.

@stappersg
Copy link
Member

stappersg commented Dec 5, 2024 via email

@RICCIARDI-Adrien
Copy link
Contributor

Hi ! Let's have a look !

@Neustradamus
Copy link
Member

Dear all,

Before 2.20, it will be nice if you can help to solve some Issues/PRs:

Recall, some weeks ago, I have done a comment here:

@agowa, @AdSchellevis, @fichtner, @marjohn56, @bmeirellesRJ, @mhanor, @stu-gott, @wcbonner, @the-j0k3r, @WenChao1Hou, @BrainSlayer, @Mile-Lile, @bnutzer, @stu-gott, @MarioVerbelen, @ruben-herold, @ipilcher, @pyk1998, @romanrm, @corvus1, @ihipop

If you know some IPv6 specialists, please come here, there are several tickets, thanks in advance.

@Neustradamus Neustradamus pinned this issue Dec 6, 2024
@Neustradamus
Copy link
Member

Neustradamus commented Dec 8, 2024

Dear all,

What do you think about the changes from RFC8319: Support for Adjustable Maximum Router Lifetimes per Link:

Updates to RFC 4861

In source:

@zenczykowski
Copy link
Author

zenczykowski commented Dec 8, 2024

https://datatracker.ietf.org/doc/html/rfc8319#section-3

This section makes unrealistic assumptions about multicast packet loss.
Or rather they're reasonable for wired networks, but most users' networks are (at least partially) wireless. True wired only networks, like those in datacenters are often statically configured for reliability (and thus don't run any ipv6 ra daemons). As such we should definitely do a good job of supporting wireless networks.

Due to a desire for power savings many (most?) battery powered devices (like phones, tablets) have a wifi dtim multiplier > 1 (possibly even as high as 9) which results in multicast/broadcast loss being as high as 90% (though such devices basically just cannot be made to work well with ipv6). [Note: the dtim multiplier usually also depends on the wifi AP dtim interval, but far too many APs use a value of 1 for the interval, instead of using 3 or even 6 - likely due to bad defaults in hostapd]

More reasonable devices top out at a dtim multiplier of just m=2, and suffer from 'only' (m-1) / m = 50% multicast (and thus unsolicited RA) loss.

As such K needs to be not 3, but more like 15+. All lifetimes in the RA should be at least K times the MaxRtrAdvInterval.

For power savings reasons you also don't want (any) lifetimes to be below 30 minutes, and ideally closer to 2-3 hours (or more).

As such wrt. https://datatracker.ietf.org/doc/html/rfc8319#section-4:

In Section 4.2, inside the paragraph that defines Router Lifetime, change 9000 to 65535 seconds.

sure.

In Section 6.2.1, inside the paragraph that defines MaxRtrAdvInterval, change 1800 to 65535 seconds.

doesn't hurt, but not at all useful.

In Section 6.2.1, inside the paragraph that defines AdvDefaultLifetime, change 9000 to 65535 seconds.

sure.

But the real changes should be to the K=3 multipliers, ie.

MinRtrAdvInterval
Default: 0.33 * MaxRtrAdvInterval If
^ this should not be 1/3, this should be 1/15 (or even less)

AdvDefaultLifetime
Default: 3 * MaxRtrAdvInterval
^ this should not be 3*, this should be 15* (or even more)

@heliosfa
Copy link

heliosfa commented Dec 8, 2024

This section makes unrealistic assumptions about multicast packet loss.
Or rather they're reasonable for wired networks, but most users' networks are (at least partially) wireless.

I'm not sure it does make unreasonable assumptions at all, given modern networking practices. Multicast on 802.11 networks happens at the minimum data rate of the network, which can potentially be set to sub 10 Mbps. This means it's usually faster to do multicast-to-unicast conversion for WiFi, and many access points do this by default. Whether they do this for IPv6 RAs or not is very much down to implementation...

RFC9119 has some discussion on the topic.

@romanrm
Copy link

romanrm commented Dec 8, 2024

Before 2.20, it will be nice if you can help to solve some Issues/PRs

In my view they asked to just tag 2.20 to solidify the current state and have already implemented features get into distros (as there has not been a release in 3 years). It is not feasible to expect to "quickly" solve all the outstanding issues collected over all that time before making the release. Else this can take another 3 years easily.

@melroy89
Copy link
Contributor

melroy89 commented Dec 8, 2024

It is not feasible to expect to "quickly" solve all the outstanding issues collected over all that time before making the release. Else this can take another 3 years easily.

I actually agree here. Normally, it's better to take the more "Agile" approach, by releasing more often with smaller releases, and thus smaller iterations. Instead of trying to release one big release every 5 year. This also ensures that you receive timely feedback, which you could include in the new release.

@Neustradamus
Copy link
Member

Neustradamus commented Dec 8, 2024

@fichtner
Copy link
Contributor

fichtner commented Dec 9, 2024

FWIW, did a bit of testing in OPNsense/FreeBSD and the current development state seems fine. Found just a small thing in the form of a warning during compile #242

@stappersg
Copy link
Member

stappersg commented Dec 9, 2024 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

No branches or pull requests