Mr. Webpage

Cutting the mustard

responsivenews:

The browser is a hostile development environment and supporting a wide range of desktop browsers can be tough work.

One of the immediate challenges we discovered when we first started the responsive news prototype was the large range of devices that we would have to support. It terrified us. This article is about a solution we use to alleviate this problem.

Read More

hg update

After spending the last 7 or so years in subversion, the switch to mercurial is non-trivial. There’s a lot to un-learn (I still type svn sometimes). 

All the books and tutorials that I’ve read keep making these claims about how mercurial is so easy. Mother-uckers doth protest too much. Sure, the syntax is nice and coherent. But conceptually, in a team environment, mercurial is way more complicated.

In “traditional” client/server version tracking systems, you have exactly two things to think about. Your working copy, and the server’s definitive copy. 2. It’s a nice small number. Easy to think about.

In distributed systems like mercurial, there is no server, except when there is, but it’s just one more copy. I say one more, because the preferred method of working on stuff seems to be making local copies of the repository. Like, lots of them.

So you end up with a bunch of repositories that are similar. They are technically unrelated, but they’re kind of related. So there ends up being a graph of relationships between repositories. They are easy to merge back together, because mercurial is good at merging. But repo-relationship graph just exists in your head, as far as I can tell*. This is the part that’s throwing me the most.

The concepts do have a lot of power, I can see that. I’m just frustrated that it’s been 4 days and I’m not the god Mercury yet. I’ll get there eventually.

* when you clone, mercurial keeps track of where you cloned from. But I’m not sure how I would know that without looking in mercurial’s private .hg directory.

Python + Postgres + Lion = PITA

I have to join the apple developer program just so that I can use python with postgres? That seems a bit silly.

Installing XCode, which is free from the app store, but only after you give them your CC details. And it still doesn’t seem to be enough to get llvm-gcc-4.2. But psycopg2 requires llvm-gcc-4.2. 

Apparently, to get this command line compiler I have to open up xcode prefs and ask for it. A step that requires a full developer program account. Which requires that you validate your CC info and go through a ridiculous multi step form, re-entering your apple id about 12 times on the way.

Then you have to sign up for which track you’re going to be on. I told them I was doing safari work, because it was the only free option. But I’m not. Safari can bite me. 

I may be showing my age here, but command line tools like a compilers should be part of the standard development environment. All the gui crap is the extra. When did we start flying backwards?

CSS selector search/replace for html editors

Why doesn’t it exist? Quick, somebody, make an eclipse plugin. (and by ‘quick’ I obviously mean start a committee and work on it for a couple years. Gooooooo eclipse!)