Become a fan of Slashdot on Facebook


Forgot your password?
Get HideMyAss! VPN, PC Mag's Top 10 VPNs of 2016 for 55% off for a Limited Time ×

Comment Unicode (Score 1) 200

Maybe the Unicode consortium need to supplement the ZWJ character with an SJW character.

In all seriousness, while Unicode is intended to be universal, I don't think that bullshit emojis have a place in it. We have markup in addition to characters, not to mention SVG and other formats which can better represent them. They don't have to be in the Unicode standard.

Comment Re:KDE is the Premire Linux Desktop. (Score 1) 111

Nothing special. It was a fresh clean install in both cases. It's even a problem without the NFS home which has older crufty bits in the config. I've seen it on both my work machine and home machine, with Intel Iris and Radeon R9 390 graphics, respectively. The NFS problems are due to the KDE programs and libraries liking to log verbosely to stderr which gets logged to the homedir. While it's not as noticeable with local discs, it becomes a significant problem with freezes lasting up to a minute with NFS. This is a major usability problem. There's really no need to log this useless junk--this has been a problem since KDE4.

Comment Re:KDE is the Premire Linux Desktop. (Score 1) 111

For me kwin crashes for no reason very regularly (10 mins on average); back to i3 for me since Ubuntu 15.10; 16.04 is no better. It also completely sucks when using an NFS home. Every mouse click seems to want to write some pointless logs or state back to my homedir, freezing everything up. I don't have that problem with i3: it just manages windows and doesn't do lots of unnecessary I/O. Still pining for the pinnacle of usability and stability which was KDE3.

Comment Re:It is not like it was not expected... (Score 0) 251

Most people inside Britain don't find him funny either! I've never been a particular fan, and I don't know of anyone who is. I can only imagine he's connected with the BBC management, and is suitably PC. I doubt Rowan Atkinson would have the needed personality either. He's not a presenter, he's an actor, and is known for acting roles rather than being himself. One major change with this new series is that they hired actors rather than car nuts who can talk well. The old trio might have scripted parts of their show, but their banter came across as quite natural, and they all knew their stuff even if they didn't focus on it too much. They weren't acting, they were doing stuff with their mates, while being filmed doing it. I think LeBlanc could grow into that, and it would help if he had some more compatible personalities which complemented each other. Evans definitely wasn't that person! And regarding the "non-PC" attitudes of the old presenters, mentioned elsewhere in the comments. People watch Clarkson because he's an opinionated arsehole, or rather an entertaining and funny opinionated arsehole. It wasn't the sanitised and bland product which is the rest of the BBC's output, and his views likely resonate more with people than they care to imagine. Hell, I have plenty of friends in India and other countries who think Top Gear is the best thing on TV and absolutely *love* Clarkson. This is another demographic the BBC would like to pretend doesn't exist (they should be outraged by his non-PC opinions!), and while they all think he's an arsehole (as do we all), outraged they are not.

Comment Re: Compression (Score 3, Insightful) 295

It depends entirely upon the type of data. See the examples below. Some data gives a little compression; there's a lot of binary data in my homedir which doesn't compress too well. On the other hand, source code, system logs and mail can compress superbly.

Regarding the "performance penalty", it's generally going to be positive and improve performance. We are talking lz4 here, not gzip/bzip2/xz. It's fast, trading a lower compression ratio for performance. It can compress and decompress blocks in parallel. It can do this much faster than it can read data from disk, so you'll actually improve read and write speeds. And this is on top of ZFS being able to pull data of multiple spindles as the data is distributed over multiple zvols, with redundant copies of data, etc. It's likely not the penalty you think it is. It does multiple rounds of lightweight lz4 compression to reduce the entropy, and it bails out early if poorly compressible.

% zfs get refcompressratio red/home/rleigh system/usr/ports system/usr/src system/var/log system/var/mail
red/home/rleigh refcompressratio 1.33x -
system/usr/ports refcompressratio 1.60x -
system/usr/src refcompressratio 2.16x -
system/var/log refcompressratio 6.54x -
system/var/mail refcompressratio 4.99x -

With compression like this, you no longer need to bother compressing rotated logs. And while the homedir compression is small in comparison, it's gained me an extra 100GiB just for this single dataset, which is not to be sneezed at.

Comment Re:Good Luck (Score 1) 295

The licence is most certainly compatible with BSD. It's included in FreeBSD base for crying out loud!

% uname -srm
FreeBSD 10.3-RELEASE-p4 amd64
% find /boot -name '*zfs*'
% find /usr/src/ -name '*zfs*' | wc -l

Comment Re:Compression (Score 1) 295

As jedidiah stated, by putting it in the filesystem it's completely transparent. No program reading or writing needs to care; it just works. To add to that, it's not "one size fits all". With ZFS, you can tune it on a per-dataset basis. Here's a (shortened) sample from my NAS. Whitespace stripping due to slashdot, not me.

% zfs get compression
red compression off default
red/data compression lz4 local
red/data@20150615 compression - -
red/data@20150724 compression - -
red/home compression lz4 local
red/home/rleigh compression lz4 inherited from red/home
red/home/rleigh@20150616 compression - -
red/home/rleigh@20150724 compression - -

As you can see here, lz4 compression is enabled on some datasets, but the default at the root of the pool is off. You aren't limited to lz4, there are other compression options. And in reality, it's much more granular than the dataset level--this is just the default for new files being created in the dataset.

Comment Re: Compression (Score 1) 295

It doesn't assume that. It's not compressing the whole file as a linear stream like it would with gzip or similar; it's chunked by compressing per-block with repeated rounds of lz4 (or whatever algorithm you picked). It will seek to the start of the first chunk containing the start offset you requested, and start decompressing from there. It works just fine and is completely transparent, including memory mapping and everything. It'll likely retain the whole uncompressed block in the ARC until you write it back or discard it.

Comment Re:IPv6 is a failed technology (Score 2) 112

The growth curve is clearly showing exponential growth here, and we're now well into the rapid adoption phase. Yes, the absolute value is 11% (now 12). but it will continue to grow with increasing speed. It *is* coming. It took a while, but it's a juggernaught which can't be stopped now. We'll all be using it in a couple of years at this rate. All the major ISPs have committed to do this, and network effects will drag the rest along in time.

I've been on native v6 for three years now.

Comment Re: Why, or why not ZFS? (Score 1) 150

No. Because it doesn't violate it. And being a monolithic blob is the least of the criticisms which we could make about systemd, when there's an entire book's worth of bad design in there. ZFS was designed by competent and expert professionals, rather than unprofessional prima donnas, and it shows.

It's a fundamentally different design to traditional UNIX filesystems and disk management, but that doesn't automatically make it a monolithic blob. Is Linux LVM a monolithic blob? That's the level your question is at, as well as being flamebait.

Internally, ZFS is layered similarly to a Linux raid/lvm/filesystem setup. Here, you would have raw block devices managed by hardware or software RAID, with LVM using these devices as physical volumes. It would then provide logical volumes upon which you could create filesystems.

With ZFS, you would have block devices aggregated into "vdevs", which would be the equivalent of RAID0/1/5/6 RAID sets. These are the equivalent of LVM physical volumes. Next, you would use one or more vdevs to create a "zpool", which would be the equivalent of an LVM volume group. Finally, you would create datasets in the pool, which are the equivalent of a logical volume plus a filesystem, or a zvol which is the equivalent of a logical volume--a raw block device. So it's cleanly and logically layered. It's using plain block devices as the backing store as for any UNIX filesystem, but it's not creating intermediate block devices as LVM does--it's managing that internally.

The layering is pretty much the same--it's a well separated design. What's different is that ZFS has knowledge of all the layers and can use that to do things much more efficiently and much more robustly. For example, when doing a RAID rebuild ("resilver") it only needs to resync bits of the disk that actually have data on which can dramatically reduce the statistical likelihood of encountering an unrecoverable error. A dumb RAID setup doesn't know that, and will fail if it encounters an error during a full rebuild; ZFS will succeed if those errors were in areas weren't in use. And it can also be instructed to keep more than one copy of data for important stuff, which gives it an even higher chance of rebuilding in the face of corruption. There are a whole pile of other benefits as well, but as an admin the main benefit is that it's a dream to manage on a day to day basis, and you can even delegate management of sub-datasets to other users and groups, so they can snapshot their own data at will, send and recv data, create new datasets etc. The design is clean, well thought out and brings features which are completely missing from anything else.

Slashdot Top Deals

"Confound these ancestors.... They've stolen our best ideas!" - Ben Jonson