"Startups, comment gérer une équipe de développeurs" par Laurent Cerveau

Binary Crosswordshttp://www.laurentcerveau.org

[email protected] [email protected]

• Last time we talked about hiring an engineering team

• No we talk about creating a product with this team

• I do not pretend to give a top notch description of all technics mentioned here.

• In fact this is even be the opposite. This is not at all the purpose of the talk

• So again a manager oriented thing :-)

We start where we left

• You have ideas/specs/people (3)

• You ask around for all the best tool and technologies…a little bit

• You still want to believe technical development is deterministic

• You have read a book on how it works

• Your planning is so that you will get investor with a SkyRocket stuff in 6 months

Ready to make a product?

“Life is what happens when you’re busy planning other plans”

So after approximately 6 weeks

• You regret to have left your job to start a company

• You hesitate at learning all programming languages yourself: “I know it should be done this way”

• Process tools have already changed twice

• One guy has left, so you hesitate to fire a second one, because 3-2 = 1

Somewhere this is disturbing• The more you talk to so called “experts”, the

more you are confused

• You know where you want to go, and what your product should feel like.

• Simply you do not know which road to take, where to start fix it, and you need more than an anchor

• Time to step back

So let’s find our way

• Apprehend the technical complexity of a product of today

• Tune your setup

• Consider the constituting blocks, and what neuronal connections should be developed.

It’s so easy for your kids

Preamb: again, the following are based on my personal feelings along this path. I do not pretend to write an history

book.

A little bit of history

Dawn of personal computing

Unix MacintoshZX81

Micro Computing: we write applications to manipulate data and first of all to create documents (text, music, …) and move them

to an another support (CD, paper). Content is the output and we forget Unix

1970 1981 1984

A networks comes in

Server comes back : we write content to present it and technology to search it. Clients are strong but start to be seen as way to browse content, in addition to

create it. Age of “universal hope” : all Microsoft, all Java, all HTML

1993 1995 1998/91995

Mosaic Windows 95 Altavista Google

Free of my movements and friends

gmail iPhone

Mobile development is creating a stronger link client to server. Browsing of document turns to service (API). Social media provides influence of other. Cloud computing bring programmable computing power. Document export is now barely an issue (press online players,

Streaming music)

2004 2008 2008

Facebook

2006

AWS creation

My data, my objects

2009 2010 2013

Big Data (Hadoop) Nest Founded Android market lead

Mobile becomes the number one priority. Technologies are created to handle massive data processing, connected object appears. We have gone so far from document creation. All this has become a full graph

and power consumption is a concern.

Complexity

Desktop computer

Mobile

Server

Connected objects

Complexity

Complexity

Complexity

Complexity

Paradox• The more complex it is, the easier it looks. Think

about how works a human body. Marvelous machine still not imitated

• In vast complex system, each small components brings its own world and complexity. A little bit like “all cells contains the DNA”

• A saying goes that the problems solvable by kids are the most complex one in computer science

Do not mix the small complexity versus the bigger vision one

Additionnally• Natural movement: when one has understood

something, it becomes a reference. The more you deal with technology, the less you believe in one, and the more you think about the usefulness of all. Techno of today fades out tomorrow

• Creativity brings creativity. Why do some people still look for the universal programming language or method or whatever, when the global system is about individuality. Paradox with above.

Additionnally (2)• The more you advance in development the more

the simplicity of your product will translate into the proper use of complex technology

• Said differently : trying to force a complex system to fit into one simple view is like putting a circle in a square shape (of the same surface)

If you believe there is one “easy” solution you’re doomed ,After a while it is as much a question of feeling

than knowledge

Set up

Cards in hand

Humans

Tools

Means

Process

Cards in hand• Humans are your team. You should enthusiast

them every day, while driving strongly into one direction

• Means are all the technologies available. There are so many and understanding one is already a complex thing

• Tools are here to help you accomplish your goal. But they also require a learning curve and will do nothing by themselves

Process will transform all those elements in a fluid system

Humans

The CTO• A respectable CTO should be a at first a

choreographer

• A respectable CTO should still code with what he doesn’t know and knows how to debug.

• A respectable CTO should always have the question in my mind : “If I was really coming to this team today, how much time would it take for me to get in, and what would I understand in regards to the target goal”

Do not believe a CTO is only someone with “more technical knowledge”

The team• As a macro system/ as a micro system

• Macro should always prevail for a product development.Enforcing this will prevent individualism

• Each ones brings competence and a Dreyfus level model to the macro system.

• Creating a grammar for the product in your team is essential

In the end, written form will prevail. The one who believes that it is evident because it looks like will lose.

Process- Engineering Project Manager

• EPM is a concept I met in a big company.

• Job description is “remove the road blocks”

• Takes note, nag people, manage the bug stack, do report. This is a no-bullshit role.

• Think about him/her as a second right hand

The only process that will work is the one that you work hard enough to transform into an habit. Someone needs to work

harder

Common pitfalls• Complexity is always done in the name of simplicity.

• After development freedom comes too much process

• “Do it now” works better in 75% of the cases. Otherwise write a bug

• Why are people so afraid to throw away things in the name of “in case of”. Throw them

• Avoid arguing, enforce rules

• A so called expert is having only his own brain

The advancing guitarist

By Mick Goodrick

“The best guitar book ever written””

Apply to software development

The advancing guitarist for product development

• Explore the power of limitations

• Don’t be too fast: spend time on all what you can get from a detail

• If you enter competition keep it for you. Works fine within your team. Being alone when you work, is the case in 90% of the situations

• Nothing is ever done. This is not the case in guitar learning, why would you believe it in Software development

Means

• The vast fields of available technologies

• Stuff of today will be partially obsolete tomorrow. Whatever people pretends. Still fundamentals stay. Each generation makes a problem important

• Choice is always a problem. Take the one that corresponds to your company need : recruiting, time to market, cost

Means

Can I develop faster with this one?

As a pure personal opinion, and based on my experience, I consider no technologies brings faster development in

absolute

But…• If you get class A developers, it will go faster

• Some technology will be faster to reach an intermediate state (e.g do I need a prototype)

• Some technologies will get an ecosystem momentum.

• There are always cases where a technology will fit your constraints (e.g databases)

• You may choose a technology for a part of the way and switch to another. Be contextual.

Tools

Rule of thumb

• The only good tool is the one you will adopt. No other. You often read “The tool does this”. This is not true: you do this

• You need basically

• a bug system ,

• a build/distribution system,

• Ideally a test and/or report system

Bug system

Some example I did cross

It has to become the center point. This is the rule to enforce “write a bug”. Defer the philosophical

discussion on “is it a bug or a feature”

Anyway…it’s a clever todo list…

Build and Deployment system• Make software available as soon as it is done.

Everyone has the right to some regression

• Goes with creating different environments: prod, dev.

• Ensure the fact that “anyone” can deploy. If mistakes are made this will be a good lesson

• May be painful on mobile (provisioning profiles…)

• Goes also with a source version control system and with versioning

Better more versions, even not finished, than too less

Test and report system

• Testing has a value only if you test something that you believe will not happen

• Testing comes through usage. It is understanding how things should work

• Automate tests before a big operation : warning this is time consuming

• Reporting is way to link to the user: you may need to create it yourself

Example: mobile client reporting• Create a log system : defines event class ,

format, storage mechanism.

• First have user send it manually through email

• Format the email nicely

• Make sure it can be activated remotely

• Make an API to report it

• Have it work when no user is registered

Each step will help you understand what you track

From the start report

So what?

• The tool created here provides a common reference, a common vocabulary and talks to everyone

• It is a preamble to real user situation and so is not an “internal use tool only”

• It makes the team ask questions about what happens really in regards to what is supposed to happen

• It get cold facts

Digging into test and usage is about being drastically honest

Some Technical verbiage

Goal of this part

• Provides beacon, push all questions around one topic

• Give some familiar technology names

• Certainly not be exhaustive

Elements in the complexity

Infrastructure

Server

Mobile Desktop

Elements in the complexity

API

Language

Misc

Storage

Infrastructure

• The power engine of your system: compute, storage, network

• You’ve heard about cloud: infrastructure becomes programmable and instantly configurable: one more element gets the complexity DNA

• Main Providers

What are we talking about?

Many other players around

Answer to particular need, with different level of abstractions and easiness of manipulations. Complete ecosystem around

• Your system will have to be accessed, back-up, monitored, restored (in emergency), started and stopped potentially by many developers, secured, code deployed

• Contrary to the belief that it makes infrastructure easier, after a while it forces you to understand many concepts of networking

• With great power comes great responsibilities (this is a personal advertisement for Nephorider)

• The eternal question : it does not work the same on my local machine : Docker!

Implications

Languages

The short/no discussion version

• There is assembly

• There is C

• There is Lisp

• OK… maybe Java

Maybe it requires more: explore this

• Procedural, Object-oriented, functional

• Compiled, just in time compilation, interpreted

• Statically typed, dynamically typed

• On the client, on the server

A language provides a way to think about a problem

Servers

What server are we talking about?

• Web server versus application server

• Web server deals with the requests

• Application servers do the work

• Between them there is something called CGI (roughly).”I pass the work to my neighbor”. When it fails find which one.

Web servers

• The first in line: needs to be fast

• Multiple server for fixed constants and fast answer: CDN

• Handle protocol. Remember 80 (non secure) and 443 (non secure) : verify you talk to the right guy, speak in coded language. Read hackers book

Application servers

• May or not work in a CGI like setup (PHP)

• They bring a methodology to develop application (MVC)

• They link endpoints request to objects, concepts and storage

API

Scope of word: application programming interface

• Exposes a nice learnable and usable model with a set of functions of methods of the concept of your product

• Apply also to the libraries and frameworks you are using

• Not an HTTP only concern of course!!!

• Should never exposes how it is implemented

Some principles

• Make sure it is learnable easily if you want it to be adopted

• Make sure it is 200% coherent

• Use the HTTP vocabulary available, read about REST concepts

• Make it auto describing : no ambiguity when reading…or parsing. Uniquely identify each entity

Words you may bumped into

• Slices (things little by little), fields (get only partial elements in answer)

• JSON means nothing

• XML neither…

Storage

Storage for what?

• Organize and keep data “that has a meaning”

• Stored data has another form than live one. It may requires extra data to access it fast

• Always question of compromises : how much size to store it, how fast can you access it, how fast can you write it

• Growth of the system is important (sharding)

Databases, databases• At first we all wanted to think structurally

• SQL is a programming language designed for managing data held in a non relational database manegement system.

• We can say this is talking about a model

The NoSQL movement• Different types all addressing a particular

problem in the SQL world or dealing with particular type of data, in some memory configuration

• Do not use it only to “do NOSQL”. When you use it , adopt the mindset

Document Graph

Key-Value Column based (Big data oriented)

Desktop

One of the most famous technology trio

Looks so easy it becomes incredibly complex and powerful

• Originally Document based

• Separation of content and display

• Document became programmable and turned into applications

• Integrate in an ecosystem : tags are element of a semantic puzzle (e.g SEO)

Evolutions

• Full feature frameworks (AngularJS, EmberJS…) provides methodology to develop application

• Graphics become more elaborate (e.g SVG)

• Handle different output device (e.g Bootstrap) this is what is called “responsive”

Strangely enough I find it has less profusion of “new” technologies…but this is very personal

Mobile (native)

Lanscape

Send back to the past

• Back to some “forgotten” problems

• Power consumption, network quality, speed of execution

• But way more demanding as “output level” : usability, animation…

Communication device

• Not only as phone

• But as a set of sensors for me (e.g GPS…)

• 2 paths communications: normal HTTP, Notifications in all ways (APN…)

Misc

• Search : use component (ElasticSearch) or Service (Algolia)

• Social networks : gives at first a way to identify yourself, but also access to a vast database of data/users (OpenGraph)

• Measure your result : Analytics

A few more things to thing about

Thank You!