On 1/31/19 2:52 PM, Jim E Bonfiglio wrote:
Hi Jeff- have data been gathered regarding these file systems which the "vast majority of users" use? If so, where might I find these data? Does SUSE or anyone else gather these data regarding users of (open)SUSE? If so, how are these data gathered?
Best, Jim
On Thu, 2019-01-31 at 14:43 -0500, Jeff Mahoney wrote:
On 1/31/19 1:42 PM, Jim E Bonfiglio wrote:
Hi Jim- so far as I'm aware, thinking does not necessarily match reality. I couldn't tell you how many attend that conference, but I do wonder what a representative sample of (open)SUSE users would report as "necessary" file systems. Without such a sample this proposed change appears quite reckless.
Best, Jim
On Thu, 2019-01-31 at 16:45 +0000, Jim Henderson wrote:
On Thu, 31 Jan 2019 04:37:08 -0500, Felix Miata wrote:
And yet there are enough to manage to have an annual convention: http://www.warpstock.org/
I don't have a horse in this race, but going to the site for the Berlin 2018 conference doesn't make me think that this affects thousands of users. It makes me think it affects about 20. Small conference rooms, a relatively small lunch gathering. Makes me wonder what the actual attendance is.
The solution is simple - remove the filesystem from the blacklist. Seems like a really easy thing to do for something that's used by 20 people.
It's not like the filesystem module is not being built or included.
In terms of local file systems, the vast majority of users use: ext2/3/4, xfs, btrfs, and reiserfs[1]. OCFS2 and GFS2 are common enough that I wouldn't want to blacklist them. JFS saw some use in the past but it was never particularly popular since it was the disk format used by OS/2 and not the AIX version. My opinions on it aren't new[2]. The rest are for compatibility with now-ancient or otherwise uncommon OSes, for pure flash (ie: without an FTL) media, or have other maintenance issues (f2fs). Are there users out there for any of these? I'm sure there are a handful. I just don't think it's worth leaving the attack surface open for the vast majority of users to accommodate a few -- especially if we can minimize the impact on the few.
Now that I look at it again, the list was composed using the file systems we build on SLE15. There are two others that should be added: omfs ntfs[3]
-Jeff
[1] I expect this will go on the blacklist at some point in the future as I'm the last person to do any real work on it and I have other priorities that take precedence. [2] https://sourceforge.net/p/jfs/mailman/jfs-discussion/thread/fpps5p$g2t$2@sat... -- I'm assuming the staff member was me in this case. Dave also says that he tries to fix bugs but the hard ones go unfixed. This was 11 years ago. [3] Not that NTFS isn't widely used; Just that the only attention that the ntfs kernel module has has seen since 2007 has been for tree-wide cleanups and API changes. Use ntfs-3g instead.
Please observe the list convention and refrain from top-posting. We don't have specific data because we don't do install tracking beyond the standard http logs which don't yield anything more concrete than the number of unique users who downloaded a package. If you'd like to propose deeper, more fine-grained statistics, you're welcome to lead that discussion. This is anecdata based on experience. I've been a contributing member of the Linux file system community for 19 years and have led the SUSE kernel file system team for the last 12. All software has bugs and the lack of bug reports for e.g. befs doesn't mean that it's perfect. It means so few people use it that they're not hitting any. Mailing list activity may also be decent metric of activity and popularity. The btrfs list had 12600 message last year and 897 messages since 1 January 2019. The XFS last had 10200 message last year and 571 since the start of 2019. I don't keep an archive of the ext4 list, but some screen scraping says about 4700 messages last year and 330 message this year so far. I wouldn't use these numbers to /rank/ popularity, but they're sufficient to show that the communities are still active. In contrast, the JFS list had 229 messages last year, most of which were courtesy CC's for tree-wide API changes. ReiserFS had even fewer, but it was our default file system for several years (and, as I said, it'll be added eventually). In terms of commit activity, qnx6 has been largely untouched since it's initial commit in 2012. The last time the qnx4 maintainer committed code he wrote was in 2009 and that was to remove write support entirely. All changes since then have been janitor-style changes. The last substantial changes to adfs were in 2011. I know HFS is unmaintained because I saw a CVE relating to it show up this year and everyone shrugged. The list reflects reality and the blacklist is the most user-friendly solution to shrink the attack surface while allowing flexibility of deployment. -Jeff -- Jeff Mahoney SUSE Labs -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org