[OpenIndiana-discuss] Merging OI + OmniOS? (And OpenZFS vs ZFS)

[Jim Klimov]

On January 8, 2021 1:27:34 AM UTC, Chris <oidev at bsdos.info> wrote:
>On 2021-01-07 13:50, Guenther Alka wrote:
>> In the end I do not think in these categories.
>> 
>> A few years ago all my own servers were OpenIndiana and maybe 40% of
>my 
>> users on
>> Solaris. Now I am 100% on OmniOS beside a OI evaluation machine (and
>miss OI
>> features and use cases) and maybe 10% of my users are left on Solaris
>with 
>> OI near
>> to not relevant now.
>> 
>> OmniOS and OI are OpenSource. Why not take over this from OmniOS to
>keep 
>> them more in sync?
>> On Linux Debian and Ubuntu are big enough to be independent while
>very 
>> similar. On
>> Illumos everyone may be too small alone in the long run.
>IMHO No that RHEL is in bed with IBM, and as a result; RHEL, and CentOS
>
>taking
>quite a different path. OI is likely to see a greater influx of users.
>A 
>quick
>look on the CentOS mailinglists/forums indicates thier userbase is
>LIVID 
>about
>the changes. The was not one single positive note about the change.
>tl;dr
>A lot of CentOS, and as a result, RHEL users; are abandoning ship.
>They'll 
>all
>be going *somewhere*. Maybe it's to OI. :-)
>
>--Chris
>> 
>> Gea
>> 
>>> 
>>> What you want is to make a OmniOS extra respoitory (git) with GUi
>packages. 
>>> That would factually mean abandoning OI and the Goal of a General
>Purpose 
>>> OS. That would be an acquisition not a merger.
>>> 
>>> -Till
>>> 
>>> _______________________________________________
>>> openindiana-discuss mailing list
>>> openindiana-discuss at openindiana.org
>>> https://openindiana.org/mailman/listinfo/openindiana-discuss
>> 
>> _______________________________________________
>> openindiana-discuss mailing list
>> openindiana-discuss at openindiana.org
>> https://openindiana.org/mailman/listinfo/openindiana-discuss
>
>_______________________________________________
>openindiana-discuss mailing list
>openindiana-discuss at openindiana.org
>https://openindiana.org/mailman/listinfo/openindiana-discuss

Alas, I'd not be too enthusiastic about people leaving CentOS because of rolling-release model coming to OI because it is the primary model for OI Hipster too (with half-yearly snapshots). Maybe OmniOS LTS would be a better fit for them in fact - but still, if that debacle provides an influx to illumos ecosystem, that's good :)

Elsewhere, it was assumed that omnios is "close to" illumos-gate. In fact, it is not, and in a way stages the improvements that take time or have other hiccups to get into the requirements for common upstream codebase changes.

There are several facets to this. One is getting those fixes into illumos-gate - that is hard for the same reasons those commits are not upstreamed yet. Another is people who missed features like lx, bhyve, overlays, etc. that are/were absent in OI and present in OO. It might be alleviated by building and installing a gate (and additional packages for features) made from OO codebase. Tribblix has a flavor like that. Such work would probably need some integration effort, but probably is not prohibitively hard. It may oppose current goals of Hipster as a distro made on top of vanilla illumos-gate however, so there is a question of who'd make it and keep it working.

Yet another aspect is that build systems and packaging approaches and layout into package publishers (repos) are way too different. It may be possible to emulate one build system by another in terms of making recipes for equivalent result, but scrap that other in case of distro-merging. IMHO from sheer size and integrated magic for dependency checks and build rituals, the oi-userland would survive such merge, but not sure how kindly the OmniOS community and maintainers would take to that. After all, by goal and design OO is a minimal-footprint distro for people who custom-build their application software (with OO style of recipes or somehow else); similar to Gentoo if I were to draw parallels.

Initially I was among those who proposed such a merger when OmniTI stepped down, but no longer push for that. Not objected if that "just happens to converge", but forcing that to happen either. As tools with their certain purposes, the distros are too different to easily mash and still meet all existing use-cases. As communities of real people with different goals for their different works and projects, they often share the same pool of talent. And it is nice to have different points of view and ways of solving problems, just like shadow compilations expose different bugs to let us make more robust code.

Possibly a facet project where full OI userland builds on top of omnios-gate (plus code for packaged tools of the new features) instead of illumos-gate might solve the practical user woes. IMHO as far as OI is concerned, someone outside maintains some *-gate content and quality, so it should not be too relevant *which* third-party code is used for this or that variant build.

Jim

--
Typos courtesy of K-9 Mail on my Android