A new version of the Joel Test | Hacker News Reader

A new version of the Joel Test

startuplessonslearned.com

54 pointsprogga19 comments

"I know plenty of people who prefer more advanced source control system [than Subversion], but my belief is that many agile practices diminish the importance of advanced features like branching."

The author instantly lost any credibility with me after this line.

Why? I thought some companies do not do branching? They have their own processes?

Some went nuts with the whole turn "on/off" features (so no branching).

I thought some team within Google also don't do branching? Always HEAD?

We use git at my workplace without branching, and we practice continuous deployment. The advantage is that you spend less time merging, and you ship things in shorter intervals and smaller increments, an important part of CD. Some of us do use local branching, but we don't push to remote branches.

Facebook doesn't branch, all of their code is written against trunk.

is this true? I'd like to hear more about this? How can all code be committed to trunk? Don't they need to branch for releasing at least?

Oh, yeah, they branch for releasing. That is true.

John Carmack recently said he implements alternatives "in parallel" so he can compare them, instead of by branching http://news.ycombinator.com/item?id=3266741

[my definition: branching is non-code ad hoc compile-time polymorphism at an arbitrary set of points]

That line stood out to me too, since I have found the opposite to be true. Sounds like it's coming from someone who's never seen the benefits of switching to a DVCS.

He was quoting verbatim Joel's original list from the year 2000. I don't think the author (in 2008) needed to expound on what appeared/changed/matured in the interim.

The original list didn't mention anything about branching: http://www.joelonsoftware.com/articles/fog0000000043.html.

I agree that agile development eliminates some of the needs for branching, but I think there are still many useful ways to add it into your workflow.

You are quite right about the original list -- I unfortunately cannot edit my comment at this point.

Now the article has lost some credibility with me. I agree with you about branching. Although technically what the article says about branching is true, I don't think someone who sees branching the way you and I do would write it that way. (The article didn't lose all credibility since it's from 2008, when although DVCS were reasonably mature, they weren't pervasive.)

Agreed. With git (and I suspect this is probably inherent to dvcs) branching is very basic functionality, not an advanced feature.

Just how would (non local) branching be useful to someone writing a web application or doing client projects?

I can understand branching being useful for a local copy so that one may implement a variety of features and then picking and choosing which to integrate.

But barring special circumstances, there is only one version of a website out there running in the world, hence branching isn't at all that useful.

While reading this, it's good to keep another article of Joel Spolsky in mind: http://www.joelonsoftware.com/articles/FiveWorlds.html

Some of the author's "updates" to the test rules probably depend on which of these worlds you live in.

The different "Worlds" are deinfitely true; However, I've been in embedded SW for almost 5 years and it's becoming clear that these worlds are all starting to bleed together.

An example is the Television I work on have about 50% of their SW that can be updated from the internet 25% that can be updated if a "tech" goes to a customers house and opens the set... and 25% that may or may not be upgradeable depending on X, Y, Z, the alignment of the planets, and the country you live in...???

However, TV manufacturers are beginning to use more web content and web technologies to develop look and feel for various features as well as using third party web based APIs to power large chunks of the new functionality.

So the article says all items on the Joel Test have stood the test of time except three: schedules, specs, and testers. These are supposed to be superseded by agile practices. I'm skeptical. For one, agile methodologies like XP and Scrum had already existed for years before Joel wrote his article.

This is from 2008, please add something like [2008] to the title here on hacker news.

One key point to keep in mind is that this article is specifically speaking about startups, not mature organizations, so obviously some of the modifications to the original test and added points won't be universally applicable.

I think the best point Eric Reis adds to the original test is the last point in the article: "Do your programmers understand the product they are building and how it relates to your company's strategy?" Such a crucial point in startups, especially when they start to grow quickly.

I think this test is rigged to make Joel's company look good (no offense, I love his writing and style and it is probably his honest opinion that these are the best ways to approach software development).

Reply on news.ycombinator.com