Comment by hinkley
2 years ago
I thought the usual solution when we disagree irreconcilably with a maintainer was to fork and rename? I always thought it a bit weird that we were naming parts of Linux after people. We already did that once and that kinda uses up your freebie.
Seems like everyone quit Hans and nobody rallied the project back together.
I don't think Reiser's name had anything to do with the removal. If I recall the situation was more or less as follows:
Reiserfs3 was out there, and while it had plenty fans (good performance and efficient disk usage), it had a fair amount of corruption issues. Reiserfs4 was about to come out. Reiser insisted that Reiserfs3 was done and an obsolete relic, and reiserfs4 was the new hot thing.
Reiser ran a consulting company of some kind. IMO this may have put a bit of a damper on contributions.
He also had a contentious personality, with quite a lot of people disliking him, and him having trouble convincing people to merge reiserfs4.
Right about that time the whole murder mess happened. So reiserfs3 was apparently abandoned, reiserfs4 was uncertain if it was going to get merged. Namesys, Reiser's company of course fell apart and the existing employees had to find something else to do.
So that was probably about the worst timing possible. Reiserfs4 didn't get merged, Reiser was dealing with the trial/prison, and other filesystems started showing up as well.
> I thought the usual solution when we disagree irreconcilably with a maintainer was to fork and rename?
Yes. But that requires someone, or a group, to take responsibility for that and support the fork. Maintaining a filesystem can be a complex undertaking.
> Seems like everyone quit Hans and nobody rallied the project back together.
I think it was more like he was the core of the project with others contributing. Once he was out of the picture no one else had sufficient passion and/or time for it to take on the mantle of project lead sufficiently (to push Reiser4 onward and eventually getting it merged into the mainline Kernel and maintaining Reiser3 in the meantime & further forward).
While Reiser4 is still maintained, it has never been merged into the mainline kernel limiting its support in common Linux distributions. I don't know if that is because the current maintainers have tried to have it merged and failed for some reason, or if they have not pushed of its inclusion at all.
What is deprecated and due to be removed is Reiser3, which is not actively maintained. There are some technical issues that would need addressing soon if it were to remain, and in any case an unmaintained filesystem is a dangerous thing to rely upon if you can avoid doing so. It isn't being removed because of who started it, it is being removed because it is not well enough supported for mainstream safety.
Reiser3 won't be removed until some time in 2025, and unless you need the latest latest kernel at all times an active setup will keep working for a while after that (until the older kernel it uses falls into EOL), so you have plenty of time to migrate if you need to.
If a lot of people were relying on Reiser3 there would be a lot more noise about this. People using Reiser4 are building their own modules (or patching a kernel tree and building it in) already and this will not affect them.
If I remember correctly (and I followed the discussions on LKML at the time, but it was a long time ago), Reiser4 was not merged because it had some "cool" features that could cause problems.
For example (again, IIRC), it allowed directories to have hard links. Al Viro was adamantly against it, showing that this potentially creates some very serious problems. In particular, it allows for cycles in the directory graph (it's no more a tree), and Viro has shown that detecting or preventing such cycles may by prohibitively expensive, and undetected they would cause any program that does directory walking (like search) to loop infinitely.
This was not the only problematic feature, just the one I remember more vividly (in part, because of Al Viro's caustic argument style :-) ).
I didn't follow LKML directly, but from what I saw elsewhere that seems right. I remember there being technical arguments against Reiser4, and the discussions about them getting interesting (not combative, but very direct) in part because Hans wasn't one with a "nicey nicey" discussion style either!
Though until Hans was out of the picture, those discussions were still on-going.
Remember that XFS was added to the kernel in 2001, and was already well-supported by most (all?) distros by the time the Reiser3/4 issue was beginning.
People who encountered issues with Reiser3 usually migrated away, and never looked back, Reiser4 was already DOA imo before all the other stuff happened.
Instead, people integrated the interesting ideas of reiserfs into ext2, and made ext3. This is also a perfectly fine way to handle disagreement.
Also, people quit him way before that thing. He mostly supported the filesystem alone.
There is just no need for such FS today. Everything is bloated, random access is much faster (SSDs, NVMe), storage is cheaper.
It's the only FS that I lost data with.