Monday, July 23, 2007

Learning the holy ways of consulting

You might've noticed another fall in the number of posts here lately, because yet again it is time for summer vacation. I've now been vacating for a week, and it feels like I've been away for a month. So, in order to keep my fingers warm, I'm semi-planning the arrangment for 9 new Java-developers we're receiving in the beginning of August.

The reason I dug into this job of training the newly recruited is obivously that I want to turn these new peeps into miniature versions of myself. They're (mostly) fresh out of university, are all smart and pleasant people (I met them on a couple of previous occasions), and seem like good potential mini-me's ;)

Jokes aside, I have a strong opinion on how new guys and gals should receive the most basic knowledge and in what order this should happen. Personally I was quite abrubtly launched onto a customer's site, but luckily I had a couple of years of experience in the right direction, so I already had basic stuff like agile practices, IDEs and Subversion-tools in my finger tips.

Now before I go on about how my job is best taught, I have to ref this blog-post by Scott Adams. Basically, he asks his readers to describe their job in one sentence. If I had to describe my kind of consulting, it would be:

Figuring stuff out.

We figure out how stuff works. Tools, frameworks, techonolgies, products and code. Then we teach this stuff to the client. A guy I know in uni recently stated "I would like a job where I get to test a bunch of different frameworks and try lotsa different stuff in a bunch of different ways..". Well welcome to the company, buddy, because that's exactly what consulting is about (well, my kind of consulting anyway. You have another kind of consulting which involves pushing crappy software onto clients, but in a way that's a good thing for us because there are lots of clients willing to hire us to clean up afterwards when these other kind of nasty consultants have made a total mess of things).

Some more advice to any new consultant starting out.

You're young and silly. This means you can get away with being completely honest. Yell out when things are going wrong. Just do it, and keep doing it for as long as you can. If you're lucky, the skill will stick and make you an even more valued consultant than.. well, what you would've been.

I like to think that I am still sincere and honest, but I'm afraid that after a while in the profession, we earn a chunk of respect and rumour around in our company, client-range, blogosphere, etc. and we want to maintain this amount of.. honour, in a way. We don't want to be admit our mistakes, because that lowers our value as developers.

The previous statement is wrong! Being able to admit your mistakes increases your value as a developer! Remember two things though:
  • Admit the mistakes before anyone else finds them.
  • Never bring along a problem (inclduding the ones you have created) without suggesting a solution.
I belive honesty is the most important value of agile development. But enough of that now, onwards with my advice to new consultants.

Learn the strokes! I saw some slides from a Dr. Heinz Kabutz talk (Productive Coder) in TSS Barcelona, and there was one slide where he advised: Spend hours learning your IDE: The keyboard shortcuts, the refactoring, the auto-completion. This will speed you up! There are only two commands in Eclipse I can not do without the mouse, and I hate it because the mouse is slow! Know your tools well, it will speed up your development, and also spread across the team as you keep giving them small hints and tricks that make their day easier. And it will make you look good/helpful, which are are important traits of a good consultant.

Be curious! If you want to get good at figuring stuff out, get going. Try to figure out stuff. Download this and that open source project and build it for yourself. Try the tools out. Set up a server, create an application, just do it. Try it. I know it is hard, most software seems pretty crappy after the 10-minute test, but there is a reason the software exists, and its probably you who haven't gotten round to find the correct solution (yes, it might be poorly documented). Keep trying, get on the mailing list, ask the community how it works, but first, read the f****** manual. I hate reading manuals (mostly I prefer tutorials), but that's where the answer will be.

Spread the knowledge! Brag about the stuff you do. Share code and write about the stuff you are doing in the company wiki. Write a blog! People will notice and respond to your engagement and activity and give you feedback when they found the stuff you wrote helpful or inspring (if you do not receive feedback, you're either doing a poor job of profiling yourself or working for the wrong company). This is also an important aspect of consulting. Inspire the people you work with.

Now, enough inspiration for one post. I will get about two days with the new guys which will be Java-development only (agile concepts included). We'll do some micro-sprints a couple of hours a piece and run through as many of the basics as we can. This post was actually going to be more about the concrete content of the sprints, but I think its long enough already. Back to summer-vacation :)

Sunday, July 08, 2007

Jazoon 2007: Some more notes on last day

I'm just back from some dubbin' in Publin, I mean pubbin' in Dublin, so little time to write a full post. However, I did find the notes I lost from day 4 in Jazoon:

...

Last Day. Neil Gafter on closures.

First talk OSGi from Ergon Informatik.

The speaker is really happy with OSGi. It's great, it works really well, etc. Running osgi demo.

Bundles/plugins managed in a phone on a console app called J9 Console.

Goes on to talk about RCP. eSWT for embedded. Subset for SWT. Can run without RCP/osgi context if you don't need them.

JFace, gui toolkit. eJFace for embedded.

eWorkbench lets you bundle simul apps running. eUpdate is the update mechanism that is in eclipse. Pretty cool, the whole update manager is run on an embedded device.

Services can be shared between apps in the same VM.

Issues they have encountered in eRCP. Useful list.

Use JFace fonts and typesets to avoid memory leaks.

Some negative experiences.. Complex bundle dependencies. The services have states to manage, and this can be tricky to learn. Plus all the usual embedded ui issues.

...

Web 3.0/semantic web Q&A session with Henry Story following the previous session that we unfortunately did not attend.. Very interesting.

...

Java and SecondLife.

Seems like SL has potential in using resources as in uri resources...

Abit about the SL script language. The scripts are only evalutated when in the SL engine, client side.

Scripts can also call http requests that the SL server will shoot off to external servers. 1 KB limit on the response. An example is a SL phone booth that shoots off SMS requests to external servers.

The 1KB limit can be workaround by having the SL object being able to respond with packets. Note that client code cannot be changed once deployed many places.

Communication is used between the client and an external server but needs to be done through a remote channel. The external server then XML-RCPs to the external server.

The java library (a port of the C# library) for utilizing SL clients is somewhat dead, appearantly. But offers good control.

...

Sort of split in half when it comes to deciding on the next talk. It's either a very interesting presentation on Jackrabbit or a talk from Netcetera on the complexity of software.

After talking with David about Jackrabbit for a little while I decided to jump in on the talk bout Build-up of artificial complexity. A talk straight to my heart.

Code grows ugly. We all know the problem, so what is the solution? Discipline and modularity are old solutions. Reduce number of (external) components is another one. I like addressing every req with an external component. His advice is be sceptical and choose the frameworks/components you *like*.

People don't like being constrained away from their personal favourite tech stacks. Limitation is dangerous, so don't restrain too much.

Will read more of this stuff. Read Greene's essay, but can't find the article on the net. Will just have to wait for the slides to come online (nudge nudge, Jazoon).