APT Rust requirement raises questions
35 points by calvin
35 points by calvin
The email could certainly have been worded more politely, especially the threat "or sunset the port", but ultimately seems like the writing was on the wall for a long time and the ports could have expected that Rust would change from a soft dependency to a hard dependency in the near future.
ultimately seems like the writing was on the wall for a long time
To add to this a bit, nothing is changing about Debian's support policies here. The affected ports were unofficial or not directly supported since 2011.
Feels like the author just wants to stir up controversy. The email was firm but polite enough, and it’s just bickering over wording of communication rather than anything actually technical.
Author here: really not. If you've read the article, you'll note that 1) declaring that other ports must be sunset on his timeline is well beyond what an individual maintainer can do, and 2) including Rust in APT depends on many other factors -- none of which were coordinated with other folks in Debian before making the declaration. Also, 3) I went into some of the difficulties that Debian currently has with building/rebuilding Rust packages. They are doing the work to improve that, but most of the folks are volunteers.
Klode is paid by Canonical to work on APT and so forth, which means that much of his work has to run through Debian. But other people doing Debian work are volunteers, and it's not particularly polite (no matter the wording) to simply declare "this core piece of the distribution now requires Rust, and everyone has to accommodate that". AFAICT Klode hasn't, for example, rolled up his sleeves to help with any of the work to bring Rust to other ports, or try to do work upstream to bring gccrs up to par.
I feel like this is, at best, a developer who is really into Rust pushing an unfunded mandate onto the rest of the project. Again, if you read the article, you'll note that another APT maintainer -- who has committed more to APT than Klode, over time but not recently -- disagreed with his reasoning, and Klode never even bothered to respond to him, at least not publicly. There is plenty that is "actually technical" at play here.
I get why someone in Klode's position would be impatient, I truly do: he's tried to push some other changes for APT to reduce his maintenance burden and gotten blocked. (I'm talking here about the old-style APT sources format.) Debian can move glacially at times, and inertia is a real problem. But I'm not sure that justifies this.
AFAICT Klode hasn't, for example, rolled up his sleeves to help with any of the work to bring Rust to other ports, or try to do work upstream to bring gccrs up to par.
That's a crazy expectation. No wonder he isn't bothering with appeasing a crowd like this.
AFAICT Klode hasn't, for example, rolled up his sleeves to help with any of the work to bring Rust to other ports, or try to do work upstream to bring gccrs up to par.
He went above and beyond as is. He could just have ignored the existence of unofficial ports of Debian. Asking him to put in work that the people who actually use these ports are unwilling or unable to do seems rather unfair. If the retro architectures can't keep up, why wouldn't they be satisfied with running a retro operating system? Giving SuperH fans a veto on what goes into Debian seems like the worst possible outcome of this.
Er, how did he go above and beyond, exactly? By making an announcement?
One of the Debian project’s reasons for using Debian, as I noted in the article, is its broad hardware support and community/democratic process. The notion that Debian should just dump a port because it’s “retro” is not in keeping with its stated and traditional values. The question isn’t so much how old or how many users, but whether there are developers doing the work to keep it updated.
If Klode wants to change that, he ought to drive that discussion rather than using control of a single package/tool to do it by fiat.
Is 6 months a too-aggressive timeline? Maybe, but at some point we have to cut bait on these ancient architectures that haven’t even been supported in over a decade
Concerning how heavy Canonical is leaning in on Rust.
APT is a better idea than the coreutils; but when something goes wrong, it's going to hurt people taking the time to integrate Rust more responsibly.