Windows-Compatible ReactOS Is Getting ReiserFS Support

[duby229]

Tail packing is great but even if you mount with `notail`, you're still going to see some benefit. That's the beauty of reiserfs tree.

Packing incurs a slight performance penalty needing more CPU. I'd tried the loopback approach on my SD card but it didn't offer any benefit performance-wise but did help with better space usage. (not to the extent reiserfs does)

Ok, thanks for the info. I started google searching on the matter. I want to take some time today to do some benchmarks. I want to test ReiserFS+notail, ReiserFS+tail, ext4, ext4 on loopback, and squashfs+aufs. I'll probably use a portage tree minus distfiles as a test filesystem.

EDIT: Anybody know of a reasonably large filesystem with tons of tiny files that would be a good benchmark?

[rabcor]

Well that's cool I guess, but ReactOS remains completely irrelevant to everything and everyone except it's developers. Sadly. I mean I want to see it succeed but, it's only gone from nowhere to middle of nowhere between it's inception and version 0.4 over how many years? Oh that's right, 18 fucking years.

Still waiting for it to get somewhere; but definitely not holding my breath considering that for 18 years they still haven't even got Windows XP level compatibility (despite having a 3 year head start on it, over 14-15 years after said head start they still haven't gotten anywhere near it), they don't even have USB support yet, do they?

As cool as reiserfs support is, it is completely irrelevant to the goals of ReactOS... it should be an afterthought, you know at the point in time where at least the basic windows drivers for windows xp or 7 or 10 are actually working on it in the first place.

[GraysonPeddie]

Well that's cool I guess, but ReactOS remains completely irrelevant to everything and everyone except it's developers. Sadly. I mean I want to see it succeed but, it's only gone from nowhere to middle of nowhere between it's inception and version 0.4 over how many years? Oh that's right, 18 fucking years.

...

... +1

[W.Irrkopf]

Reiser 4 or 3.6?

I followed the link from the article to the patch. I the just grepped -i for "version" and it looks like it's version 3.6, but someone with more understanding about reiserfs may want to verify.

[duby229]

Ok, thanks for the info. I started google searching on the matter. I want to take some time today to do some benchmarks. I want to test ReiserFS+notail, ReiserFS+tail, ext4, ext4 on loopback, and squashfs+aufs. I'll probably use a portage tree minus distfiles as a test filesystem.

EDIT: Anybody know of a reasonably large filesystem with tons of tiny files that would be a good benchmark?

I ran through a bunch of benchmarks today and I've come to the conclusion that performance wise it really doesn't matter, only a few seconds difference between all of them. I couldn't get squashfs to work right with aufs, but squashfs did give the lowest disk usage followed by reiserfs+tail. And considering ease of convenience I think the reiserfs option is the one that makes better sense.

[DrYak]

Well that's cool I guess, but ReactOS remains completely irrelevant to everything and everyone except it's developers.

Uh... nope. Wine and reactos people share patches and development work for supporting userspace.

Development of ReactOS (user space) is very much relevant to everyone using Wine.

Okay maybe *today's* exemple of ReiserFS isn't (it concerns ReactOS own kernel which obviously can't share code with Wine)

And also would ext4 on a loopback image accomplish the same thing since it's encapsulated in a single file?

I can't follow the logic of this question...

[duby229]

I can't follow the logic of this question...

I was just asking because I wasn't sure. It turns out it didn't actually make very much difference.

[Azpegath]

Is it something that SSD can't fix ?

I used it for that specific purpose and it was great; for a while. Then the fragmentation happened. ReiserFS fragments very much, especially when replacing all small files regurarly, and there is no other way of defragging it than to copy to other partition, reformat the partition and then copy back. That made me leave it for the next time I recreated my system.

[nomadewolf]

Reiser 4 or 3.6? I used 3.6 back in the days and it was actually fairly good if you had lots of small files (iirc. it was also default fs for a long time in SuSE). I think it is sad that people did not abstract the murder case from the pure tech and threw both into one pot so everbody kinda dropped Reiser 4.

Anyway, it's nice to see some Linux / Unix OS in a "windows style" environment. Makes dualboots more comfortable. Though, on the other hand, sometimes it might be good that windows can't access Linux FSs -> think of malware like those crypto things recently.

>Why is that? Something to do with block size or journal?

I guess it might well be something with the blocks / inodes. I mean, that is normally the case when you have a tradeoff with the space. It's always a matter of weighing the possible use of a filesystem into the choice of minimum administration unit size / overhead. Not enough inodes saves you some overhead but if you run out of inodes this is really nasty (happened to me recently). More inodes, smaller units, more overhead. *shrugs*
I don't know about the journals but I wouldn't expect the differences to be so huge.
At least all of them should still be better than FAT16 / FAT32 (vfat), which easily wasted 31 KiB per file on small text files in most partitions of larger size.

Agreed.
The ReiserFS has some really great tech features, and should not be dropped over the creator's personal issues.

[SystemCrasher]

Well that's cool I guess, but ReactOS remains completely irrelevant to everything and everyone except it's developers. Sadly. I mean I want to see it succeed but, it's only gone from nowhere to middle of nowhere between it's inception and version 0.4 over how many years? Oh that's right, 18 fucking years.

Lol, it is irrelevant for ReactOS devs either. These pathetic losers are using Windows to do development and VMs to put a show. Fucking lame attitude for wannabe-OS-developers. Probably that's why they do not care how their OS performs. So, 18 years later, you can't even do ReactOS development using ReactOS. It also pretty much useless for anything else. If one needs to run Windows app, it is more desirable to run it on Linux, or maybe some BSDs. At least these are going to boot with significant probability on your hardware and would not lock up or kernelpainc within next 10 minutes.

Another fancy fallacy is lack of drivers. They initially declared they'll write compatible kernel, hook up windows drivers and rock-n-roll. It never worked like this. Windows proven to be moving target and those writing drivers do not care how they would perform in ReactOS either. And since goals were set like this, they haven't bothered self about open drivers. The result? 18 years later, there're still no stinky drivers. Ironically, even BSDs wrote quite some drivers within 18 years, not to mention Linux. Isn't it ironic Windows clone is worst when it comes to hardware support? That's the only real advantage of windows, and they managed to screw it up. Failbag!!!