When you (or your programs) obtain memory directly from the operatingsystem, you pretty much invariably get memory that is filled withzero bytes. The same thing is true if you ask for fresh empty diskspace, on systems where you can do this (Unix, for example); byspecification for Unix, if you extend a file without writing data,the 'empty space' is all 0 bytes. You might wonder why this is.The answer is pretty straightforward; the operating system has toput some

As part of a thread on linux-kernel where ZFS on Linux's problemwith a recent Linux kernel change in exported symbols was brought up, Greg Kroah-Hartman wrote in part in this message :

My tolerance for ZFS is pretty non-existant. Sun explicitly did notwant their code to work on Linux, so why would we do extra work to gettheir code to work properly?

If one frames the issue this way, my answer would be that in today'sworld, Sun (now Oracle

A couple of years ago I wrote about the harm of ZFS not beingGPL-compatible , which was that this kept ZFS frombeing bundled into most Linux distributions. License compatibilityis both a legal and a social thing ,and the social side is quite clear; most people who matter considerZFS's CDDL license to be incompatible with the kernel. However, itturns out that there is another issue and another side of this thatI didn't realize back at the time. This issue surfaced recentlywith the

It isn't news to me that dealing with spam makes me irritated andangry. I resent the intrusion into my email, and then I resent thetime I spend dealing with it, and in fact I resent its very existence.This is not a rational irritation and hatred; I viscerally dislikespam and people and organizations who spam me. Sensible people wouldresent spammers only for the time and effort they take to deal with,but I am angry all out of proportion with that.

(This anger is part

Many years ago I wrote about some things that irritated me aboutPython 2's socket module . One of mycomplaints was that many instances of socket.error should actuallybe IOError or OSError instead of a separate type, because that'swhat they really were. Today I was reading AdamW’s DebuggingAdventures: Python 3 Porting 201 ( via ), where I discovered in a passingmention that in Python 3, socket.error is a deprecated alias of OSError .

(Well, from Python 3

For years I've used a custom-compiled Firefox, with variouspersonal modifications . Usually this worksokay and I basically don't notice any difference between my versionand the official version except that the branding is a bit different(and since I build from the development tree, I'm usually effectivelya Firefox version or two ahead). However, I've now run into a newdrawback, one that hadn't even crossed my radar until recently.

The short version is that I read a

For years, I've had a custom-compiled version of Firefox with variouspersonal modifications, generally built from the current developmenttree. The number of modifications has fluctuated significantly overtime; when I first wrote about my history of custom-compiling Firefoxin this 2012 entry , it was probably my minimalpoint for modifications. These days my version has added significantlymore changes from the stock version, in larger part due to Firefox'sswitch to WebExtensions. The somewhat unfortunate thing about thisincrease in

One of the great things about blogging, especially over a long time,is that it gives you opportunities to be spectacularly wrong inpublic. (I'm actually sort of serious about this being a good thing,but that's another entry.)

Back in 2012, I wrote an entry called "The wrong way to harvestsystem-level performance stats" , inwhich I maintained that single-shot stats gathering of absolutecounters was the wrong way and that the host agent should computeand

Christopher Mills' IOCCC 2018 entry ( via ) is a stunning tour deforce that, well, let me quote the author's introduction:

The program delivered here is both a full machine emulation of theoriginal PDP-7 that Ken Thompson used to write the first version ofUNIX and a full machine emulation of the PDP-11/40 used by subsequentUNIXes. The Makefile can build versions that can run each of thefollowing:

1. UNIX v0 for the PDP-7 (circa