Mold 1.7 Released But May Need To Change Software License If Funding Not Secured

[ssokolow]

Rui Ueyama is unfortunately considering a software license change where Mold use would continue to be free/open-source for personal use but would not be the case for corporate users.

That's a self-contradictory statement. The Open Source Definition, Debian Free Software Guidelines, and Four Software Freedoms all reject licenses which discriminate based on fields of endeavour.

Also, just in case, I'll be `git clone`-ing the repo now.

[MauganRa]

No, the m68k work was done because me and some other community members threw some money at it. Step up, people!

Yeah, I had a hunch that it might be useful for many hobbyists. On the other hand, in its long history it had so many embedded uses that I could not imagine 68k support to *not* be commercially important.

[nanonyme]

Yeah, I dunno if he realizes this, but AGPL code is banned completely at google. I imagine lots of other companies too who might be familiar with supporting open source work. I can't even conceive of why you'd put the AGPL on a linker which is not connected to the internet. What is the expectation? That someone wold offer "linking as a service" over the web!?

Or did he just slap on the strictest, most in-your-face licence the FSF ever shat out, without considering for a moment how radioactive it is?

I was thinking this exact same thing since very inception of the license. Why exactly AGPL? It is such a footgun if you want corporations to touch it as consumers. And for a linker it seems even less sensical. But then again, it looks like the intent was always to sell *entire* IPR of the project to a single company who would then relicense it.

[bug77]

I have suggested several people they could (and should) at least once make a donation the the creator of the software they use daily. The answer is always the same: "Why? It's free!"

It's how I go about things when they ask me if I need a license for IntelliJ Idea. "I don't need to use the Ultimate edition. You need me to use the Ultimate edition, because if I lose 10% of my productivity, you're losing 3 yearly licenses worth of cash, each month".
But people can be that dense

[fong38]

I'm actually a little surprised by the comments on here. This guy set out to solve actual problems. Rather than working on yet another generic CRUD app which no one asked for (and getting paid the big bucks for it), they're instead drastically cutting down linking times which in turn adds up and ends up probably saving a lot of money for corporations over a longer time-span. If you are benefiting from their work throw them a bone at least, they deserve it!

[satadru]

Reality check: Nobody is using Mold. Most haven't even heard about it.

If it became proprietary (as if AGPL wasn't bad enough), it would simply and completely disappear from sight.

Any key good ideas it might have are visible in the already out there AGPL code. If any corpo cared about these, they'd literally reimplement them (possibly as BSD/MIT code, possibly as proprietary code) in the toolchains they use, rather than pay for a license.

That's just how it is today.

For what it is worth, in our (admittedly niche) Chromebrew distribution, mold is now our default linker.

[kiuRISC?]

I can't even conceive of why you'd put the AGPL on a linker which is not connected to the internet. What is the expectation? That someone wold offer "linking as a service" over the web!?

Or did he just slap on the strictest, most in-your-face licence the FSF ever shat out, without considering for a moment how radioactive it is?

1) Affero clause did not come originally from the FSF, who really missed the mark not adding it to GPLv3 in 2007.

2) Before you rage about this license for a linker, how about we try to find out how popular remote build services are, often called "CI/CD" or whatever, like GitHub Actions, GitLab Executors, Travis CI, Azure DevOps Server, or multiple other server-side "trust our security stack" build farms? Those of us going as far as possible in the direction of GNU-space dont need to trust such things, but let's ask ourselves how common such methods are for building code and the security ramifications for downstream users who generally just execute binaries without checking signatures are?

Really it seems auditable remote linkers are one very small piece of the problem if the plethora of such systems have any degree of popularity.

[rene]

this new release does not even build for me anymore. Not on current x86-64, not on last years GCC, nor ARM64 - https://t2sde.org/packages/mold

[Almindor]

I say all the power to the guy. Dual licensing with corporate use restrictions should be IMO the new golden standard of open source and free software. I unironically think we should start making a new class of "free from corporate abuse" software (essentially what he wants to achieve here).

Using things for free/no profit? Go for it. Want to contribute? Go for it. Want to make moolah by using this thing? Pay up buddy. Nothing wrong with that, if anything OSS has missed this boat ages ago and it's hard to go for (just look at the artards here with all the vitriol).

[ssokolow]

I say all the power to the guy. Dual licensing with corporate use restrictions should be IMO the new golden standard of open source and free software. I unironically think we should start making a new class of "free from corporate abuse" software (essentially what he wants to achieve here).

Using things for free/no profit? Go for it. Want to contribute? Go for it. Want to make moolah by using this thing? Pay up buddy. Nothing wrong with that, if anything OSS has missed this boat ages ago and it's hard to go for (just look at the artards here with all the vitriol).

It's been tried multiple times. See, for example, the SSPL. The result? Someone forks the project from the last libre-licensed commit (eg. as Amazon did when they forked Elasticsearch and Kibana as OpenSearch and OpenSearch Dashboards) and the original gets booted from all the Linux distros as "not legally compatible".

That's a design feature that debian-legal checks for with their "tentacles of evil test" (the "What if a company buys the rights and then tries to make life hell for existing licensees?" test.)

Other examples that are purely about aversion to management without license changes would be the exodus of OpenOffice developers to form LibreOffice and the exodus of MySQL developers to form MariaDB when Oracle bought Sun Microsystems.

...and it's not even purely about corporate takeovers and the like either. The modern GCC lineage runs through a GCC fork named EGCS that came about when Stallman was too dictatorial about GCC development and eventually had to cry uncle and bless EGCS as "the new GCC".

Remember, these licenses and the tests for evaluating them were crafted by people like Eben Moglen, who used "Imagine if you had to pay license fees for each mathematical equation you wanted to use." in the talk I saw. (Seriously. How would I ever start a small business or independent consultancy without pre-emptively selling my soul to venture capital if the options were "free for non-commercial use or pay for a proprietary license"?)