| |||||||||||||
The Kernel Hacker's Bookshelf: UNIX InternalsThe Kernel Hacker's Bookshelf: UNIX InternalsPosted Sep 5, 2008 0:04 UTC (Fri) by ras (subscriber, #33059)Parent article: The Kernel Hacker's Bookshelf: UNIX Internals
I am not usually a huge fan of Val's articles, but this one is an absolute gem. It should be held out as an example for anyone doing a technical book review. As reading it you came away knowing what the book does well, and what it doesn't do so well. You also got the feeling that Val has read a few of them, and so you could give her opinions some weight. She even answered the question "why would I not be better off just looking at the SLAB source code"? (Answer: because it complex, and you won't understand the reason for the complexity without the history).
As for the book covering file systems - its the wrong time for that. The SLAB allocator is both state of the art and stable. File systems are anything but. We are going through a sea change right now. Disk drives are literally becoming too big for our existing file systems. You can't copy them in a reasonable time, and in fact can't even reasonably ask the user to wait for an fsck on boot. There are so many sectors that can't even be sure some won't silently change - with the changes undetected by the disks ECC. It won't be long before a bad block scan will take a week. There are solutions emerging - zfs, btrfs, tux3, hammer. It reminds me of the activity in the version control space no so long ago. In a few years we will know what "start of the art" is, and some will write books about it, and hopefully Val will review them. But now is not the time.
(Log in to post comments)
The Kernel Hacker's Bookshelf: UNIX Internals Posted Sep 5, 2008 2:44 UTC (Fri) by pflugstad (subscriber, #224) [Link] In a few years we will know what "start of the art" is, and some will write books about it, and hopefully Val will review them. But now is not the time. Shoot, Val will be writing them.
The Kernel Hacker's Bookshelf: UNIX Internals Posted Sep 5, 2008 7:24 UTC (Fri) by nix (subscriber, #2304) [Link]
Given what she's worked on in the past (e.g. at Sun), I think it safe to
say your tense is wrong, or at least not inclusive enough. :)
The Kernel Hacker's Bookshelf: UNIX Internals Posted Sep 5, 2008 7:18 UTC (Fri) by nix (subscriber, #2304) [Link]
There are also media with profoundly different semantics emerging: very
large 'sectors' that are faster to read than to write, that can be written a distinctly finite number of times, and with no concept of head movement or rotational latency. FSes on flash certainly will look different in ten years: I hope I'm not restricted to buying ugly VFAT-faked-by-chip at that point.
|
|||||||||||||