If pair programming feels threatening at first, you’re normal. Two developers, one keyboard… yuck. If you’re like me, you want your own space, and you get a kind of rhythm going with mid-compile checking email and stuff. Taking that away during pair programming time sounds like it would be incredibly awkward.
If you’ve ever implemented a small-scale WinForms .NET project with modest database needs, you’ve probably been tempted to keep it “simple” by using Microsoft Acess (.mdb) files.
What kind of organizations benefit from an agile approach to software development?
I was thinking of this around a year ago while working with a client who was thinking of reimplementing, for the first time, some important applications with .NET. One of the business-interface people started a planning meeting by stating:
Before we do anything else, we need to plan this whole project in detail. We need to know exactly how this is going to be done. Right now.
I don’t recall exactly how I responded out loud, but the monologue in my head went something like this:
Someone asked on LinkedIn the other day how to handle technical discussions in a Daily Scrum meeting that is ending early: do you use the free time, or defer the discussion for after the meeting?
To me, that’s an easy call. You do not have time for technical discussions in a Daily Scrum, the same way you do not have time to talk about last Sunday’s football game in traffic court. It’s not a matter of time, it’s a matter of focus and appropriateness.
This question just came up on Stack Overflow. It reflects a pretty common misunderstanding of how C-style strings are represented by char pointers in both C and C++.
Greatly condensed, it goes:
- You read some data into a std:string object. You display the contents; it’s all there.
- You invoke c_str() on that std:string, and display its contents; it’s not all there.
When taking on a new Project That Sucks, the most important thing for me to do first is to find out, in a big-picture sense, what has gone wrong. Why It Sucks comes before Making It Not Suck.
Sometimes it really is as simple as needing another bright software developer. Other times, the client can throw people at the problem but the problem is really structural: it’s set up to fail, and they should consider just dropping it. And still other times there’s a team or management dynamic that could use a push in the right direction. More →
I learned an interesting, although in retrospect somewhat obvious, technique from Michael Feathers’s great book, Working Effectively With Legacy Code.
Suppose you have a class that can’t be unit-tested in any automated way because it has side effects or requires user input. Or, in the case of my application, it drives hardware that I don’t actually have readily at hand. NUnit and the like provide input to methods and test the resulting output. How can you run tests on something where the “output” is the motion of a sensor arm, or the “input” is a stream of weather data?
The short answer is that you really can’t, but that shouldn’t stop you from setting up unit tests.
Last week, I mentioned my theory that most development teams have one person who just isn’t that great at getting things done. Like spotting the mark at a poker table, you want to figure out quickly who that team member is. In poker, it’s so you can take advantage of that person. In software development, it’s so you can set a kind of intuitive trust level: something between will this person make it harder to accomplish our goal? and can I count on this person’s tasks being properly fulfilled? with a little bit of how should I adapt my work style? in there. More →
Last night I regaled The Smartest Guy I Know, as well as drsweetie’s high school pal Tony, with tales of how Kids These Days Don’t Know Anything and Nobody Does Things Right. It was a scene to warm the bitter soul of any software curmudgeon.
“So there I was,” I began, “optimizing the heck out of an ASP.NET application…” And my audience groaned–although I’m still not sure whether it was because they know how the story simply had to end, or because I was the teller and they couldn’t be sure that it would ever end. Because my stories get to be like that sometimes.
Yesterday, I finished up my part of a web development project, working with a team put together by a local software consulting company. Here are a few things I learned or had reinforced:
Figure out what the client really needs
A lot of times I pick up a project that’s a real mess, and the client thinks they need a supercoder to get it all done. I don’t think of myself as a supercoder, and even if I were such a person, the lack of one is usually not the problem. More often, when a software project isn’t going well it’s for reasons like these:
- redoing the same work
- poor choice of platform
- unclear goals
- not enough time allocated
- multiple agendas
- poor controls
- wrong kind of discipline
- reliance on heroics
- technical “bankruptcy” (when the code gets too complex to maintain)
What do these things have in common? They aren’t solved by supercoders.
You know what though? As I found it, this project had acquired good management and it had a very clear goal. Some of the specific requirements could have been better defined, but there was enough wiggle room that we weren’t really hampered by any vagueness. And our team lead, someone I’d happily worked with in the past, was very capable.
I could have applied some of my amazingly insightful make-your-project-not-suck magic, but no: what this project needed was a supercoder, or two or three. So I did the best I could in that regard.
I’m not talking about the “duct-tape programmer” type that Joel Spolsky lauds in his recent blog post. This website already had enough duct tape! It just would have been extra-awesome if I’d been able to understand the whole code base in a flash, and figure out how all the duct tape worked, rather than having to piece my mental model together one defect at a time.
Gauge the effort
In business terms, a key element for any independent consultant or small contracting firm is knowing how many technical people need to be on the project, and for how long. Nobody wants to be so overbooked they fail to deliver on commitments. Nobody wants unanticipated slack time.
At first, they told me it was a one-month “burst” kind of project. Then, based on burn rate, they scaled it back to three weeks. A bit later the client found more funding to make it five weeks. Then another burn-rate-related adjustment reduced my role to four weeks.
Does that sound awful? Maybe? It’s not. When it’s one of those park-yourself-in-a-cube-and-join-in endeavors, you rarely get that level of communication and adjustment. A lot of clients make it really hard to plan ahead. This one kept adjusting and communicating their time expectations, without being overwhelming about it. Cool.
Know who the chump is
They tell me it’s a truism in poker: There’s one chump at every table; you want to figure out who the chump is. Corollary: if you can’t identify the chump, it’s you.
The blunt truth is that in team software projects there is very often at least one person who hasn’t the skills to contribute much. That’s not always a bad thing: it can be a great opportunity for mentoring via pair programming, code reviews, or simply talking about how code gets done.
I really liked and respected the .NET developers I worked with on this project. On the downside, I couldn’t figure out who the chump was. I made my best effort, but I suspected that my process of understanding the existing code base was just taking too long. At the end, I realized that it took a month to figure out where everything was in this modestly complex system, which is about normal. Still, I hate that feeling: “Am I the chump? Really?”
One of the cool things about working at the client’s site is that it puts you in a geographic location that you can leverage. One of my Twitter pals works right upstairs from our war room. Another good friend works afternoons and evenings in the same office park. Our quality lead was a vague acquiantance from a project I did last year when he worked for a different organization.
So I got to optimize my lunch and snack time. I found out about new jobs opening up, in unrelated areas, for which other friends might qualify. I was able to refer my Twitter pal to two industry peers she didn’t already know. I got to hang out with, and get the latest relationship drama from, my office-park neighbor.
Those things wouldn’t have happened had I stayed in my home office or remained virtually connected via coffeehouse WiFi. It was a lot more driving than I like to do, and my car actually suffered, but at least there was that social benefit.
I’m just saying
It was a good engagement. The client is doing a lot better now than before we started, and their confidence will improve even more when the remainder of the team gets cracking on load testing and optimization. Congratulations all around!