Code Kills - Latest Comments

Re: http://blog.codekills.net/2012/04/08/adventures-in-x509-the-utterly-ignored-nameconstraints

David Wolever — Tue, 10 Apr 2012 13:16:47 -0000

Name constraints are supposed to limit the certificates which a CA trusted to sign. Ex: "I only trust this CA to sign certificates for ‘.acme.com’”.

Re: http://blog.codekills.net/2012/04/08/adventures-in-x509-the-utterly-ignored-nameconstraints

singpolyma — Mon, 09 Apr 2012 09:49:18 -0000

What are name constraints supposed to do?

Also: never trust X509-based anything :)

Re: Code Kills : Using git for Backup is Asking for Pain

David Wolever — Sun, 04 Mar 2012 12:51:50 -0000

I… May or may not… have my nethack saves in a Mercurial repository…

Re: Code Kills : Using git for Backup is Asking for Pain

David Wolever — Sun, 04 Mar 2012 12:50:45 -0000

Fair enough - if the files you're storing are "simple" blobs-of-bytes-on-disk (and occasionally symlinks), and it's acceptable to lose file metadata like ownership and permissions, then version control systems would be entirely adequate (and likely simpler to implement, too).
However, for a general purpose backup, those things are important (for example, consider /etc/).

Re: Code Kills : Using git for Backup is Asking for Pain

Guest — Sun, 04 Mar 2012 02:17:47 -0000

+1 for GITting savegames :)

Re: Code Kills : Using git for Backup is Asking for Pain

DavidGuetta — Sun, 04 Mar 2012 02:16:34 -0000

Actually, I'm using git for my (personal/company) backup for years. Git is one of the most intelligent filesystems already done (content-based) and is simply perfect for the task. I will not throw away the dozens of advantages of git just for the sake of damn permissions. BTW: When the worst occurs, nobody would give a damn penny for file permissions. Everybody will just want their files.

Re: http://blog.codekills.net/2008/08/27/why-i-don't-like-git

LukasEder — Sat, 03 Mar 2012 16:31:48 -0000

Yeah Felipe. "THE TUTORIAL". No way on earth is any tutorial ever gonna explain these things on a single page in a way that any sane person will get it... :)

Re: http://blog.codekills.net/2008/08/27/why-i-don't-like-git

LukasEder — Sat, 03 Mar 2012 16:30:50 -0000

I share your feelings so much!! Heck the first time I thought I might as well just edit a file in my git workspace and dcommit it and push it to GitHub... Well think again. That funny checkin will always be applied on top of every other checkin rebased from SVN thereafter. How to get rid of it? Hmm, revert it again on Git, check it in on SVN, rebase, push. Done. Whew. Screw U, Git!!

Funny guy too, this Felipe Contreras. "Read the tutorial". My ass. What IS the right way after all? It doesn't say in any tutorial that the right way is to be doomed no matter what you do

Re: Code Kills : Vim functions to change tab treatment

ysim — Thu, 01 Mar 2012 19:55:38 -0000

Ohhh so this is what happened to all the the .php files I started... :P

Re: Code Kills : Vim functions to change tab treatment

David Wolever — Mon, 23 Jan 2012 00:50:20 -0000

Sure, there are plenty of ways this could be taken care of on a stable code base. But I edit one-off files often enough that these functions are useful.

Re: Code Kills : Vim functions to change tab treatment

Guest — Sun, 22 Jan 2012 23:27:03 -0000

That would require modifying the entire codebase if the people contributing to it don't use vim.

Re: Code Kills : Vim functions to change tab treatment

Guest — Sun, 22 Jan 2012 22:22:28 -0000

wouldn't it be easier to just use modelines in a comment?

Re: http://blog.codekills.net/2012/01/18/in-case-of-wikimergency

Eva Wong — Wed, 18 Jan 2012 12:50:51 -0000

Good stuff!

Re: Code Kills : Which SSD You Should Buy

Andrew Westberg — Wed, 23 Nov 2011 08:58:09 -0000

This is just anecdotal evidence, but I'm on my 3rd RMA'd Intel 600gb SSD in less than a year. They are good about replacing them though.

Re: Code Kills : On Startups and Identity

Jason Cohen — Wed, 19 Oct 2011 18:47:40 -0000

Thanks for the mention. Agreed that nowadays it's weird not to say.

20 years ago it was weird TO say. You were supposed to be lifeless and formal. Otherwise you're not to be trusted. In software we'd say "shareware."

Maybe people are still behind the times -- both founders and consumers. But the future is clear.

Re: Code Kills : On Startups and Identity

David Wolever — Tue, 18 Oct 2011 16:38:20 -0000

I wouldn't go quite that far… But, yes, I can't help but getting a little bit of that feeling.

Well, that, or the startup is only intended to be a stepping-stone to bigger things (like getting acquired), at which point it will be shut down.

Re: Code Kills : On Startups and Identity

Harald Koch — Tue, 18 Oct 2011 16:26:03 -0000

I would automatically categorize any start-up without this information as a get-rich-quick scheme - typically the reason that people are not listed is because they don't want you to research their previous endeavours... :)

Re: Code Kills : The evils of `except:`

David Wolever — Thu, 29 Sep 2011 17:08:10 -0000

Eehh… There are still some cases where I've found the naked except makes the code cleaner… But, ya — with with + finally, there aren't many of them.

Re: Code Kills : The evils of `except:`

Mike Conley — Thu, 29 Sep 2011 17:03:51 -0000

Hey David,

I would even go so far as to say that a naked except for closing/freeing resources is no longer necessary, thanks to Python 2.5's "with/as" stuff. RIIA is golden when it's done right. :)

-Mike

Re: http://blog.codekills.net/2011/09/26/checking-types-in-python

David Warde-Farley — Tue, 27 Sep 2011 23:00:50 -0000

The way NumPy usually handles this is by calling `asarray()` or `asanyarray()` on its input. This allows you to pass nested lists, etc. in lieu of arrays, and simply coerce to the right type. Further diligence is required if you're expecting a floating point type and want to handle the case where someone gives you a boolean or integer array, but coercion seems to be the sanest response there too.

In general, I've found certain kinds of "generic code" strategies to be a boon rather than a hindrance to my numerical computing needs, i.e. writing code that works on arrays of *any* rank. I often find myself switching between representations of data as e.g. (1000, 27, 4) or (1000, 108) depending on which is convenient, but much of my code only cares about manipulations on the first axis (incidentally representing time).

Re: http://blog.codekills.net/2011/09/26/checking-types-in-python

David Wolever — Tue, 27 Sep 2011 00:50:48 -0000

You make a very good point. Thanks nfomon.

I come from an application programming background where the kind of error you suggest — input that seems to work actually causes horrible problems — is less common, so I didn't consider it.

Can you give an example of such a situation? The only one I can think of is accidentally using an `int` as part of a division on Py2.X without `from __future__ import division`…

Re: http://blog.codekills.net/2011/09/26/checking-types-in-python

nfomon — Tue, 27 Sep 2011 00:16:15 -0000

Coming from a numerical-computing background, I reeeally dislike duck-typing. I want the author of a routine to be responsible for what happens when it is called. Generic code is nice where it's relevant but focusing on it can be a type of "premature optimization".

"let the caller use your documentation and Python's helpful
stack traces and debugging facilities to figure out what they did wrong." -- This ignores the very horrible case, where the code accepts inappropriate input, performs a computation, and returns a result that is garbage, without any warning.

Re: Code Kills : Lies, More Lies and Python Packaging Documentation on `package_data`

Von Welch — Thu, 04 Aug 2011 00:34:57 -0000

I think it's worse than this. From what I can tell MANIFEST.in is used by sdist and package_data is used by bdist and install. So if you want coverages for sdist, bdist and install you have to specify your files in both places.

Re: Code Kills : Using git for Backup is Asking for Pain

Chris Howie — Fri, 29 Jul 2011 20:27:51 -0000

For backups of an entire system, yes. Personally, I *love* using Git for backups and synchronization of data across multiple computers, even for such things as save game data! Being able to "git clone" my Terraria worlds onto a new computer is really nifty.

Re: Code Kills : Lies, More Lies and Python Packaging Documentation on `package_data`

Vincent Bernat — Sun, 24 Jul 2011 07:29:26 -0000

Thanks for your post. It saves me some time.