Hacker News Reader

Comment by toddpersen | Hacker News Reader

The funny thing to me, which I haven't seen anyone else mention yet, is that way, way back in the old days, Grafana started out as a fork of Kibana. Go take a look at the first commit in the Grafana source: https://github.com/grafana/grafana/commit/75d03fc49ab4f95ee4...

Anyone want to guess what the third commit was? That's right, it was an Apache 2 license: https://github.com/grafana/grafana/commits/master?after=2bb7...

Before there was any code, there was a permissive license. If the choice had been made 8 years ago to make Kibana less-permissively licensed, it's very likely Grafana wouldn't exist right now.

And for historical fun, here's the first commit from Torkel after the fork from Kibana v3: https://github.com/grafana/grafana/commit/50e42c8bdd2099552d...

View on HN

they talk about Kibana in their Q&A:

https://grafana.com/blog/2021/04/20/qa-with-our-ceo-on-relic...

Relevant section copypasted below:

> If Kibana had been an AGPL project, would Grafana even exist? Are we being hypocritical?

> I asked Torkel, and he said that If Kibana had been an AGPL project, Grafana would likely have been AGPL from day one.

> This is of course a hypothetical question, so it’s difficult to determine how that would have affected subsequent business decisions the company might have made.

> Making business decisions to increase the likelihood of commercial success of our company while also supporting the health of our thriving Grafana community on the basis of new information and an evolving situation is not hypocritical.

IMHO, his answer doesn't address the GP’s point; that the original, permissive license inherited from Kibana arguably enabled the growth & survival of Grafana.

But the point is that they acknowledge that they forked Kibana and why. It’s no different than if someone forks Apache code to form a proprietary software company. The permissive license contributed to the success of many companies. They further acknowledge that Apache based Grafana can be forked too, but they’re betting on themselves to continue to succeed long term by focusing on good software, branding and community. The GP’s point seemed to indicate that “no one” acknowledged the Kibana origins when in fact Grafana themselves have.

You're right - I hadn't seen that in the Q&A, so I appreciate you pointing it out.

Listen, at the end of the day, it's their legal right to choose the licensing strategy that they think is right for their business, shareholders, employees, customers, peace of mind, etc. I feel like asking if it's hypocritical is a bit of a straw man, though. It's not hypocritical per se - I just feel that it's a bit disingenuous to say "hey, we have this large, successful business to protect" without also openly acknowledging that it may never have existed without the tailwinds of an era where the open source community embraced permissive licenses. Simply dismissing the question as "hypothetical" is what I take issue with.

Also, I see your shiny, new account and feel it's likely that you are really just here to defend Grafana. I get it. I don't think this change is a bad idea, necessarily. I just want some of these licensing shenanigans to be more brutally honest.

> IMHO, his answer doesn't address the GP’s point; that the original, permissive license inherited from Kibana arguably enabled the growth & survival of Grafana.

Why's that? If Kibana had been AGPL and Grafana had been AGPL from day 1, how would the growth and survival of Grafana have been any different?

Would Torkel have decided to fork Kibana in the first place if it were AGPL?

Would the same level of community have formed around Grafana?

Would it have received as many contributions and support for new backends?

Would Raj have wanted to acquire Torkel + Grafana for the original Raintank company if it were an AGPL project?

Would they have been able to as successfully build and monetize a Grafana Enterprise product with the requirements of AGPL in place?

It seems naive to argue that none of those would have been impacted by a more restrictive license.

The Grafana Enterprise point is the most salient here. The answer is no, they wouldn't have been able to create closed source plugins for a commercial product. Only AGPL code. As I've stated before: copyleft licenses are an evolutionary dead end: https://www.influxdata.com/blog/copyleft-and-community-licen....

P.S. Hey Todd ;)

Oh, hey! Nice to have at least one person here that isn’t going to jump on the AGPL bandwagon. ;)

Are you sure you're not asking questions whose answer depends on one's ideological pre-disposition?

Many companies ban agpl code, this came from legal and is taken seriously by business. Ops guy says it won't touch anything else, business says don care no.

I think I'm right in saying that, providing you fork from any commit before the introduction of the AGPL license, you can continue with Apache-licensed Grafana. Sounds like the way forward.

You mean when you would have forked before the new licensing became effective?

They amount to the same thing. A git repo is an historical document. All commits (and forks) up to the one changing to the new license "going forward" are available under the original license.

What do you mean? Any free software licence allows forks, including AGPL.

Yeah, that's pretty speculative to assume what would have been, but forking and improving/shifting direction of kibana would have been definitively possible from a license POV.

Their FAQ even says:

> I asked Torkel, and he said that If Kibana had been an AGPL project, Grafana would likely have been AGPL from day one.

https://grafana.com/blog/2021/04/20/qa-with-our-ceo-on-relic...

i wonder how much of that original Kibana code remains today (my guess is not very much), and if it were originally AGPL, would a completely rewritten codebase still be subject to it?

The license talk about derived work. And even if every line of code has been changed, it is still a derived work as it was progressively changed. So pattern and desings from the original are still in use. The Ship of Theseus is still the Ship of Theseus.

> The Ship of Theseus is still the Ship of Theseus.

you state this as if it's a settled fact (it isn't).

If you take a Shakespeare novel and replace every word in it with another word, and then remove and add many chapters to it, then shuffle all the pages, is it still a Shakespeare work or even recognizable as being derived from one? I think most would say "no".

In the case of Ship of Theseus, if you replaced every part with a different looking part, removed 30% of the parts entirely and added another 300% parts such that the function was expanded by 1,000%, would anyone recognize it as the same thing?

There has to be _some_ threshold beyond which you cannot reasonably claim the work is derivative.

Humour me, why must there be such a threshold?

If version B is modified from version A as a derivative work (therefore licensed as A) and version C is modified from version B as a derivative work of B, then would C not still be licensed same as B, and therefore A?

It seems it depends on how you view the accumulation of differences. If your only comparison is always against version A, then yes, eventually there must be enough accumulation of differences that one could claim a work is no longer derivative. But if you consider all versions between A and yours, then suddenly its just a series of gradual changes, none of which individually cause the work to become non-derivative. The Ship of Theseus parable seems to also fit this latter interpretation more -- it is after all about replacing items one at a time.

I agree with you from a practical legal point though. My argument above seems to imply that there's only been derived work since the first cell split in two. In practice this would be a nightmare; There should be a threshold whereupon a work can qualify on its own merits.

> it is still a derived work as it was progressively changed.

I think that depends. If it was progressively changed to something completely different, then, no.

Let's say in 10 days you replaced 10% of the code, with the same amount of code from a different program by another author. During those 10 days there'd be compilation errors, and thereafter the original work and copyright would be all gone (except for in old revisions in the repo), and the copyright holder of the new program could license it however s/he wanted.

It is still likely a derived work. Looks more to me in the example it was just practised the right to make changes to the original work which one could do under the given circumstances which is normally a reciprocal license - floss or else - which means even when if 100% of all letters in all files have been changed, one could still not do more than the original license granted.

and that is normally _not_ what you wrote:

> the copyright holder [sic!] of the new program could license it however s/he wanted.

However if sole copyright holder - real, not imagined - she sure can license it - or not. Or different another day.

Reply on news.ycombinator.com