Friday, December 31, 2010

Nice quotations

Although I'm not sure I get most of them - Tim Chevalier's Quotations File:
[Satanism and Wicca] are to world religions what JavaScript is to programming languages. -- Juli Mallett

...Debugging-by-printf is a universal theme that transcends cultures and concrete syntaxes. This shit is Joseph Campbell, yo. -- Jason Reed

...real programmers often wear climbing boots to work in case a mountain should suddenly spring up in the middle of the machine room. -- anonymous
Happy new year!

Wednesday, December 29, 2010

PL Predictions for 2011

  • JSON will be the default format for new internet APIs.
  • As more people use JSON, we'll see a XML renaissance, as we - for the first time - discover that XML actually gets some things right. No really.
  • GHC will get typed type-level programming and end this highly embarrassing state of untypedness. ;)
  • We'll have a verified compiler-and-OS toolchain for compiling and running some kinds of apps. It won't be x86-based.
  • All kinds of stuff targetting JavaScript.
  • Split stacks and maybe some scheduler improvements will be shown competitive with green threads in the millions-of-threads!!1! (anti-)scenario.
I'd really like to predict some insanely great new PL for 2011...

Sunday, December 19, 2010

Quick, what does this machine do?


From Dieter Rams, ten principles for good design.

HXA just published A comment adding to Dijkstra on natural language programming:
Software must always be an ‘un-natural language’ because its (ultimate, essential) purpose is different (and particular): it is not communication, it is design. ...

When you look at software, what you see is not a language, it is a
machine.
I'll have to come back to this topic at another time, for now take it as a fine weekend inspiration.

[Update: Of course, Lev Manovich comes to mind immediately. From his Info-aesthetics:
Never before a single machine was an engine of economy -- AND the main tool for representation.


[Further riffing off on this idea reveals that our major platform for delivering software is called a (markup) language.]]

Thursday, December 16, 2010

On Git

Giles Thomas commenting on A Guide to GIT using spatial analogies:

A much simpler view IMO is to consider the Git repositories as narratives, or meta-narratives, each committer being seen as an author of one or more. This is made particularly clear in the git man pages, where a predominant concept is the distinction between closing and opening — in a sense, it promotes the use of structural deappropriation to challenge the linear view of “revisions”. As Torvalds says, “[l]inearity is fundamentally impossible”, so the repository — or at least, a branch — is interpolated into a realism that includes the “current” revision (tip or head, as you prefer) as a reality. To put it in political terms, Mercurial deconstructs Marxist socialism, while Git analyses capitalist construction.

Tuesday, December 14, 2010

HTTP request visualization


You can clearly see the handshake, slow-start ramp-up and full bandwidth phases.
Packet Flight: HTTP request @ 40X from Carlos Bueno on Vimeo. (via Russ Cox)

Saturday, December 4, 2010

objectophilia alert!

Here is our latest minicluster design. (Ron Minnich on 9fans (hat tip to Chris))



Plan 9 is in the works. This is a nice test and development platform
for software. We're going to make the design available in a way that
people can easily build their own.

Wednesday, December 1, 2010

User space? Almost certainly not.

> Should this kind of thing be done in user space?

Almost certainly not.

First off, user-space is a fragmented mess. Just from a "let's get it
done" angle, it just doesn't work. There are lots of different thing
that create new tty's, and you can't have them all fixed. Plus it
would be _way_ more code in user space than it is in kernel space.

Secondly, user-space daemons are a total mess. We've tried it many
many times, and every time the _intention_ is to make things simpler
to debug and deploy. And it almost never is. The interfaces end up
being too painful, and the "part of the code is in kernel space, part
of it is in user space" means that things just break all the time.

Finally, the whole "user space is more flexible" is just a lie. It
simply doesn't end up being true. It will be _harder_ to configure
some user-space daemon than it is to just set a flag in /sys or
whatever. The "flexibility" tends to be more a flexibility to get
things wrong than any actual advantage.

Just look at the patch in question. It's simple, it's clean, and it
"just works". Doing the same thing in user space? It would be a total
nightmare, and exactly _because_ it would be a total nightmare, the
code would never be that simple or clean.

Linus
(from LWN)

Ghosts of Unix Past

One of the most awesome collection of articles on software design is Neil Brown's

Nulla Salus Extra Ecclesiam

There is always a well-known solution to every human problem - neat, plausible, and wrong. — H. L. Mencken
In software design, there often appears to be a situation where you either get a concept exactly right, or it will just be plain wrong and mess up your entire downstream development. (Maybe sometimes there are multiple good ways to solve a problem, instead of just one, but the wrong ones will always outnumber them.)

I think a good example is the difference in lexical scoping in Scheme vs Python. For my tastes, Scheme gets it exactly right, and harvests a lot of benefits from its clean design, whereas Python is just plain wrong, and harvests a lot of pain, everywhere. [In my limited experience.]

How can we avoid making systemic design errors? One way is to stick to what works. If you're designing a new PL, and your lexical scoping is different from Scheme or Haskell, I think you have a lot of explaining to do.

To know what actually works, you need taste. To create new designs without systemic error, you need a lot of midnight oil.

Which brings us back to:
Why do you glorify doing something new and stupid, when doing good things well is what people really should be admiring. — Linus Torvalds