Friday, 23 September 2016

Evil Two, Three and Four Letter Acronyms

I propose the following three rules for Three Letter Acronyms (TLA)


  • Platinum rule: Don't use Three Letter Acronyms. Ever.

  • Golden rule 1: Only use Three Letter Acronym's if they are well-known1.

  • Golden rule 2: Don't create new a Three Letter Acronym.


One of the things we must continue to do in software development is to promote understanding. Every 5 years the number of developers doubles, so half the software developers out there have less than 5 years experience.

In our industry2, we have a huge love of Three Letter Acronyms. In fact we nearly have our own flavour of English in TLA's. And this is not surprising given the lack of typing skills during the early days of computing (that continues today!). It's a lot easier to repeatedly type TCP instead of Transmission Control Protocol. So we gain efficiency on the writing side of the communication equation.

However TLA's are a barrier for understanding for the reader. Especially for novice and intermediate reader. They are particularly evil when one needs context to understand them or when we create new ones that are local to the document being read. For example: in a telecommunications domain does IP stand for Internet Protocol or Implementation Proposal or Intellectual Property? If I load a module into a CD; is a CD a Control Device or Compact Disk?

Now consider that the reader may not be a fluent English speaker. Consider that the artefacts are written once, but read many times. Shouldn't we favour efficiency of understanding for the reader, instead of the writer?

So to promote better communication I propose the above three rules for software development teams to follow. Let me repeat that first one again. Never use Three Letter Acronyms! Use regular words.


1. Definition of "well known": If I pop the Three Letter Acronym into google and it comes up in the first 5 results.
2. We are not the only industry guilty of this.

Edited 20th February, 2017

Wednesday, 24 August 2016

Code Reviews 2 - Dysfunctional Code Reviews

We've all had them; code reviews that just aren't effective. Here are a few scenarios.

I would love to hear your feedback about other types of reviews you have encountered.

As a submitter:

The '0' feedback code review

You eagerly prepare your code, a few methods and a few new tests. Submit it to the guild. Appoint the master code reviewers. The reviewers give it a +2 a minute later, with no feedback what so ever. WFT... surely there has to be something wrong with it?

The "Trivial" details code review

You submit your code to the review board. After a wait, the results come in. The reviewer wants the variables and methods in alphabetic order. They insist on putting in all the full stops on the comments. They are making it a blocker that you don't know the difference between "it's" and "its". They point out you are missing capitalisation in your javadocs. You haven't expanded three letter acronyms on their first use…

The "goes on forever" code review

We submit out code for review. A few days later, the first set of reviewers find numerous "serious" issues. We address all the comments. It goes out for review again. The following week, the next set of reviewers find another set of serious issues. We address all those and our code goes out again. Another week goes by and next set of comments back highlight another, different set of "serious" issues. We address those and the code goes out to the reviewers again. The next set of comments back and highlighting some of the same stuff as the first review, that the third code review made you put in…

The "Self" review

You are using a code review tool because the organisation process requires you to use one. However your team has no real interest in each others work. When you make changes, you put them up for review. But you always push through your own code anyway.

As a reviewer:

The "token" code review

It's 16:40 on Friday. The sprint ends in 15 mins. Bam! Into your mail box, comes a review for a critical feature for this sprint. The change set is long, several classes with a few hundred lines of code. Just remember it has to be in the main branch by 16:45… and that includes any re-work! It only takes a fraction of a second to hit that +1 button and the weekend is yours!

The guy who is always right review

The author is never wrong. How can you criticise this guy? Their code is flawless. Always. Even when you spot a real issue, they won't listen to you. They'll justify the unnecessary, out of scope code in the solution with fancy subjective words like "extensible" and "future-proofing". Some sections of the code might be unreadable to you, but that’s only because you are a lesser programmer and we'll save several nano-seconds during execution because he used an obscure construct. Basically this code is right. You are wrong.

The guy who is always wrong review

This guy just doesn't get it. The solution proposed isn't just a little bit wrong, it's completely wrong. There are no tests whatsoever. There is no concept of the agreed style guide, he just copied a large chunk of code from anther part of the system - changed a few strings and it's the proposed solution?

The death by a 1000 updates' review

So you get the call for a code review into your inbox. You gallantly volunteer to take it on and you fire up the tool. You spend 40 minutes reviewing the problem statement, then the tests and pick your way through the solution adding your comments as you go. Next thing you check the tool and there is an update to the commit. Many of your comments are now null and void due to the changes submitted. No panic. You start picking your way through the new solution again. Thirty minutes later, in pops another set of significant changes, rendering many of your comments null and void... again! So you start the review... again!

The "rutting" code review

A senior developer has submitted their code for review. Another senior developer has taken up the challenge. These two never see eye to eye. The comments come thick and fast with slightly confrontational and inappropriate language. It's a thinly veiled "you are an idiot"! The gauntlet has been thrown down. The submitter is not going to take that sort of "threat" lying down and they respond in kind, in the tool. The insults and accusations go back and forth for all to see in plain text. The tension rises. Other reviewers fall away, afraid of getting caught in the cross fire. Sooner or later either one declares the other one is wrong or Godwins' law will apply! At this point someone else will step in, merge the code and promises to keep these two apart in future!


Updated 23rd September, 2016 Updated 3rd July, 2017

Tuesday, 23 August 2016

Cycle time and Delivery frequency

Delivery Frequency

Delivery Frequency is the count of the number of times we ship the product out of the team to a receiving organisation. That could be a test group or a "friendly" customer or another interested stakeholder.

Generally we favour higher delivery frequencies. The optimum number depends on our delivery infrastructure. This can vary greatly across products. But at a guideline, we are looking to ship more than once per day. On products with long build and test cycles, this can be a challenge. Why this is important:

  1. It's easier to integrate smaller changes, more frequently.
  2. We get earlier regression feedback on our changes.
  3. It can help reduce "traffic jams" at the end of sprint.

Cycle Time

Cycle time is defined as the length of time a story is actually being worked on. The clock starts once the story leaves status "open" and it stops once the story is marked "closed". Usually this is when the story is delivered out of the team. Why lower cycle time is important:

  1. We get an earlier demo to the product owner.
  2. Better chance of delivering something in the sprint.
  3. There is the possibility of more agility from the team. We can change new work planned before the stories are started.
  4. Small stories are easier to implement, are easier to integrate into the system and they are easier to regression test.

Relationship of Cycle time to Delivery Frequency

There is a coupling of Cycle Time to Delivery Frequency that is interesting, and observing the trends in both can be indicators on how the team is performing.

Cycle TimeDelivery FrequencyWhat it means
Low or trending downHigh or trending upTeam is heading towards high performing
Low or trending downLow or trending downTeam not delivering despite closing stories. Need to investigate why?
High or trending upHigh or trending upTeam is frequently delivering "undone" work - need to investigate why? Are stories correctly split?
High or trending upLow or trending downTeam might be struggling. We need to engage the team to discover the difficulties and challenges they are facing.

Tuesday, 9 August 2016

Code Reviews 1 - Why?

I'd like this to be part of a series of articles on code reviews. Comments and suggestions welcome. There are as many views and opinions about code reviews, as there are developers. These are my views. I hope they make logical sense. They have been formed through my own experience.

Why do we do code reviews?

There are primarily two reasons why we engage in code reviews. Both are equally important.

Code Quality Checkpoint

For any "product", a review should be part of the production process - this goes from a set of accounts to a presentation. Reviewing one's work is a great practice for any craftsman to part-take in. There are two quality aspects that code reviews can help with.

Code Quality

How can I know that the software I have written is "good"?

Code is written once. Code is read many times. Frequent code reviews help me to write great code.

Code reviews help to have a good code base in the product. A "good" code base is one that is readable by a majority of programmers, without too much intervention by the Author. This has been expressed as the number of WTF per minute. We can ensure any style standards are consistently met.

Product/Component Quality

Code reviews are the also important for finding any possible bugs or errors in the solution we have put together. Since code reviews occur early in the production lifecycle, they are one of the cheapest ways of detecting and preventing issues.

Knowledge Transfer

Code reviews are a medium of communication between developers.

When we start out, we share our work with our more experienced peers so that we can learn. As we get wiser, we share our work with our peers to get their validation that we are doing the right thing. When we are at the top of our game, so to speak, with many years of experience we also take an active part in reviews to help pass on our experience and to get exposure to possible newer ways to do things that more recently trained developers will have learned.

Friday, 29 July 2016

Risks, Assumptions, Issues and Dependencies (RAID)

This post is about a tool that is useful to have when doing Release Planning.

During Sprint or Release Planning it is useful to capture any Risks, Assumptions, Issues and Dependencies that we identify during the planning, that might impact the successful execution of the plan. A RAID diagram is useful for this.

Setup 4 flipcharts on the wall, where the team can capture Risks, Assumptions, Issues and Dependencies.

Risk

A risk is something that might happen. Should it happen it will have an adverse impact on the success of our project. Risks should be discussed openly by the team. We need to evaluate how likely they are to happen and what mitigation actions should we put in place to avoid or to counter them.

Here is some good advice for managing risk

Examples: There is a transport strike looming for the next number of weeks, so it will be difficult for people to get to the office. Peter is doing maintenance on an other high priority project; he might get pulled on a regular basis from this project if there are issues for the customer.

Assumption

An assumption is something that we think is true or likely to be true and we rely on it being true to successfully delivery our project. Assumptions should be tested as early as possible during the execution of the plan.

Examples: To cause least disruption, 21:30 is when we think we have least number of users on line so it's the best time to upgrade our system daily.

Issue

An issue is something that is causing us a significant problem today and prevents us from succeeding in our project. The main difference between Issues and Risks, is that Issues are a certainty today. Issues need to be dealt with all the time. It's important that an owner is identified and a strategy for dealing with the issue is agreed. Can we resolve the issue ourselves? Do we need outside help? Can we avoid the issue altogether?

Examples Issues: We have scalability problem after 10, 000 connections to the database. Mary, the principle tester, is going on 2 months extended leave for July and August.

Dependency

Something that must be delivered before we deliver our project. Dependencies must be managed and monitored. How are we going to make progress while we wait for this dependency? If dependencies run late, how will that effect our delivery? Can we minimise that impact?

For example: We are using a beta version of our platform for development, and it's due to be officially released 6 weeks before we go live.

Thursday, 28 July 2016

Team Building Exercise: Tennis darts game

This game is a fun game to promote team communication. Good introduction to new team mates.

This is one of a series of simple and fun team games to help team forming.

Team:

4-8 people

Aim:

to score the hightest number of points by landing the bean bag inside the scoring zone.

Equipment:

  1. Spray paint
  2. Tennis racket
  3. 6-8 mini Bean bags ( or a balls that won't bounce)

Course:

  1. Draw out a large circular target on a large area. 4m-10m in diameter.
  2. Draw some smaller circles inside the large one. At least 5.
  3. Mark out a base approx 15m from the target zone.

Scoring:

  1. Smallest circle in the middle gets 200 points
  2. Next circle gets 15
  3. Next circle gets 12
  4. Next circle gets 9
  5. next circle gets 5
  6. Outside that the score is 0.

alternative scoring: Use hula hoops placed on the ground, further away hoops get more points

Instructions:

  1. Allow one practice bat per team member.
  2. Each person takes the racket with one bean bag.
  3. They stand inside the base.
  4. They launch the bean bag towards the target on the ground, using the tennis racket.
  5. Repeat for the next 5 bean bags.
  6. Add up the scores of where the 6 bean bags landed to give the team score.

Team Building Exercise: Risky Bowling

A great game for team building, it promotes communication around risk. Do team members keep it safe and score for every roll. Or do they go for broke and try and maximise their scores, at the risk of scoring nothing.

This is one of a series of simple and fun team games to help team forming.

Aim of game:

To score the highest points score by rolling a ball to a stop between the lines on a lane. The highest scoring zone is furthest away & the smallest zone. Be careful not to over judge the distance, as this score will be 0!

Team:

2-6

Equipment:

  • Level flat surface
  • Coloured tape
  • 10m+ roll of carpet, 1m wide
  • Bowling balls or large tennis balls
  • Measuring tape

Course:

  • Lay out the carpet.
  • Mark off an area after 3m, 5m, 7.5m, 8.5m, 9.25m, 10m

Scoring:

  1. 0-3m Zone 1 - 1 points
  2. 3-5m Zone 2 - 3 points
  3. 5-7.5m Zone 3 - 5 points
  4. 8.5-9.25m Zone 4 - 8 points
  5. 9.25 - 10m Zone 5 - 30 points
  6. 10m+ Zone 6 - 0 points

Instructions:

  1. Allow one or more practice rolls per team member.
  2. From behind the start line, roll the ball along the carpet.
  3. Where the ball rests is the individual score.
  4. Repeat & add up all the scores for the team score.

Monday, 18 July 2016

Team Building Exercise: Gutter Ball Challenge

A great game for whole team interaction. Promotes teamwork, communication and lateral thinking to achieve the goal as fast as possible

This is one of a series of simple and fun team games to help team forming.

Aim: To transport as many balls from one bucket to another using gutters

Team: 4-6 people

Equipment:

  1. 2 buckets
  2. 5 x 1 - 1.5 metre lengths of PVC guttering. To make game harder, use down pipes
  3. 100 tennis balls (bucket of tennis balls)
  4. Stopwatch

Scoring:

  • 3 minutes after the clock starts.
  • 1 point per ball.
  • 200 points if all 50 balls are transported successfully

Course:

Place the bucket of balls in a space.
Place the empty bucket approx 5-20m away. (Further away, game more difficult). (10M)

Instructions:

  1. Allow one or more practice runs per team member.
  2. Start the clock.
  3. One person stands at the bucket of balls. They cannot leave the bucket and are the only one who can handle tennis balls.
  4. The remaining players take a gutter each and line up between the two empty buckets. The should stand so that end of each gutter touches.
  5. Persons who hold a bucket can only walk when there is no ball(s) in their gutter.
  6. Person at the bucket, places one ball on the first gutter.
  7. The first person passes the ball, via tilting the gutter towards the next gutter.
  8. Once the ball is passed the first gutter, the first person runs to the end of the line.
  9. Once you reach the bucket, deposit the ball into the bucket using the gutter. All return to the start.
  10. Continue until time is up.
  11. Count the number of balls transferred.

Penalty

Should the person with a ball(s) in their gutter walk or run, the ball is detucted from the total.
If a ball falls, it is lost and cannot be picked up again.

Hint:

You can try to transfer more than one ball at a time. You may have more than one line at a time. However let the teams figure this out themselves.

Thursday, 19 May 2016

I did pair programming with my colleague, do I need a peer code review?

In my opinion the short answer is: yes

Pair Programming is a tool to aid production of an artefact, be it code, a user document, an email, a presentation or some other artefact.

Peer Review is a quality assurance tool or checkpoint in our process. We are checking the output of the production process. Lack of knowledge of the production process is a benefit for Peer Review, however you must understand the problem being solved. It is a clean set of eyes over the problem and the proposed solution.

Pairing and peer review are therefore complimentary and neither one depends on doing the other. It’s a good idea to do both. We know that pairing during production produces less defects. We know that peer review helps catch defects before the customer gets the product.

What pairing means for peer reviews is that we should find less issues and for those issues we find, their solution should be less time consuming. There should be less "returning to the drawing board". But pairing doesn’t mean we should no longer do peer reviews. A good quality process demands a quality checkpoint.

Updated 15th August, 2016

Tuesday, 3 May 2016

Email communication

The single biggest problem in communication is the illusion that it has taken place. George Bernard Shaw

In this post, I propose that email amplifies the illusion of communication and it should be avoided on most occasions.

Reliable communication is one of the biggest challenges of the human race, not alone the software development workplace. Effective communication is as equally dependent on the receiver, as it is on the sender or initiator of the communication.

There are a significant proportion of our engineering fraternity who demand and insist we use email because it's become an integral tool in the office place in the last 25 years. However, contrary to popular opinion, I would like to propose that it has to be one of the worst communication mediums.

In agile we favour Individuals and Interactions over processes and tools, so at an academic level we agree that email is terrible (it's a tool!), but practically day to day, we don't live it.

I see at least the following short comings of email as a medium for communication.

Problems

Feedback: The first problem with email communication is the lack of feedback on the message we deliver. Unless a receiver actually replies, we can only assume they read the message and that they understood it. Generally you only get a swift response to an email when you made an obvious mistake or the receiver strongly disagrees with you!

No tone: Communication is not just what you say, it's also how you say it. The same words said in a certain tone, have completely the opposite meaning (sarcasm). It's very difficult for the reader to detect the tone in an email. All we have is the literal words. However, often the email sender will compose the message as if they are saying it to the receiver, in their head - tone and all.

Mistakes: We easily make mistakes in what we write or type and it's nearly impossible for us to spot our own mistakes.

  • Grammer matters, for intent. A comma missing or inserted incorrectly changes the meaning of a sentence.
  • We can easily leave out a word - this can invert the meaning
  • Fundamentally many of us can't proofread our own mails (Pair email writing?)

Temporal aspect: We can't control when the reader reads the information. With the spoken word we have instant receipt - we can watch the facial reactions of the receiver. With email we assume the reader reads our mail pretty soon after we sent it.

Speed reading: We don't read every word because we quickly read email and loose out on critical detail. Or worse we click on the mail in their mail tool and think they have already read it, but didn't at all.

Volume = Noise: In my experience, the sheer volume of email I receive daily, makes it easy for me to miss important events and information. Some tools help by filtering mails - so you need to know who's "more equal than others" and focus your time reading just those mails. But the large volume magnifies the lack of feedback, the temporal aspect and the need to speed read.

Have you anything good to say about email? It's a great historical record. Most email tools have powerful search facilities. Sometimes this is useful.

Conclusion

If it's important to land your message and it usually is; first talk to some one directly or by phone. Use normal human communication to reach a conclusion and agreement easily and quickly. Then you can follow up with a short email with the decision you agreed so that they can search for it later!

Thursday, 28 April 2016

Agile Gluttony

if you want to bring in Agile Software Development to your organisation because of:

  1. you want to go faster
  2. do more
  3. do better
  4. to save money
  5. any combination of or all of the above
then you and your organisation is suffering from a serious case of Agile Gluttony.

If your primary motivation is any of the above you should really consider incremental process improvement in your current processes. I have seen on a few occasions, the first attempts at agile means a complete tearing up of the current process and the cutting out of nearly everything, except coding.

Unfortunately you can't go faster or save money by cutting corners. You can't do more by diverting everyone to coding tasks. You can't do better by delivering more code. When inevitably the organisation fails to meet the expectations the customer puts on it, there is great disappointment everywhere. The valley of despair for everyone, including the engineers, management and the customer, is a deep one.

There is only one reason to adopt agile software development processes. That is to satisfy your customers needs by the frequent delivery of working software. You care that your customer gets value. When you adopt agile for that reason, in the long term you succeed and reap the benefits. However the costs of putting in the necessary infrastructure and around your application and the coaching around your development teams that enable rapid and frequent deliveries will cost you a significant sum.

When you are able to frequently deliver working software to your customer; you don't need to go faster because you are perceived as responsive to their needs. You typically actually do less, harnessing benefits the old 80:20 rule1. You will definitely do better because you deliver less change per release and your automated test suits catch bugs the customer had before. And you will save money by not doing the features the customer doesn't want or need.


1. The 80:20 rule states that 80% of the time a user will use 20% of an application features. The features in this 20% set, are the most important features and need to work well all the time. However we need to be aware they may not be the most valuable features. The most valuable features are the ones a customer is willing to pay us for.

Thursday, 7 January 2016

Happy new year 2016

Welcome to 2016

It's a brand new year!

Recap

2015 was a great blogging year for me. I got to put down some of my thoughts on lots of things from Java, unit-tests, test coverage and aspects of teams. I set a target to write at least one article per month and I managed to achieve that. Just like 2014, I haven't published anything in November and December, but I have been gathering ideas...

Plan

So the plan for 2016 is more of the same. I want to continue to blog about software development. I'm currently in a management role, so my topics are going to swing more towards topics on managing and leading software development teams, with a small sprinkle of technical stuff, just to keep it interesting (for me!). The target, once again, is one blog entry per month. I'll definitely expand into topics such as code review and pair programming.