�ݺ�ߣ

Balancing Sharing
And Sel?shness
The CPAN Way

Hi

I'm Michael Schwern
I went to CMU
I failed.

Comprehensive
Perl
Archive
Network

CPAN is 90% crap

90% of everything is crap
(present company excluded)
That's ok because ideally you never ?nd it
And nobody uses it (for long)

Only 10% gets used

That 10% is really good

Isn't that inef?cient?

Yes...
...but maybe it's the least inefficient in the long run

Oslo QA Hackathon

Lots of arguments and productivity
Lots of things that have been festering for years resolved

Improving CPAN
Improving Perl testing

Why Us?

Why this bunch of people?
Why not somebody else?

People use

People use our stuff

Merit-ocracy?

We're not there because we wrote the quot;bestquot; thing
What is the quot;bestquot;?
How do you select what is the quot;bestquot;?

Do-ocracy?

We're not there just because we do the work.
90% of CPAN is crap
They did, but they're not selected.

Used-ocracy

We're there because our stuff gets used.
That's it
Even if it's internals are crap, it's useful

By what people need

Not just what they think or say they need
But by what they actually need

Use is measurable

Clearly measurable
Downloads, phone homes, quot;popularity contestquot;
Mailing list activity, robust communities
Forks, patches...

Select By Use

A healthy system selects by use

CPAN is a Used-ocracy

The users select what is important.

Pave the cow paths

We were in Oslo to pave the cow paths

Descriptive

Describing
More formally
Because the community had grown too large for conventions
Word-of-mouth only goes so far
Spread best practice

English dictionaries are primarily descriptive

Proscriptive

Every time we proscribed something
Came up with our own best practices
Strayed off the cow paths
It all went to hell
We didn't have enough information
Caused implementation problems

Proscribe the potentially good

also proscribe the good

Proscribing change

Nobody will ever want to do it this way
Proscription backs you into a corner
Proscription bets everything

YOU'RE
WRONG

I'll win
So many people make so many predictions over so much time
Odds are they're wrong

Be Descriptive

Describe only as much as you need
Leave the rest open as a user playground
Even the bad things
They provide wedges to do amazing things years later

Descriptive embraces change

The world will change
Technologies change
People change
Safe bet
Have a system that works with change

Can't predict change

In the long term
Who could have predicted the web?
NCSA Mosaic 15 years ago in 1993
So you can't control it

NCSA Mosaic is 15 years old

(The stable releases)
That was the beginning of the modern web.
The ?rst graphical web browser.

Perl 5 is 15 years old

Developed before the web took off

CPAN is 12 years old

And for much of that time it wasn't intensely used.
Not like it is now.

Perl predated the web
Perl did not predict the web
Perl became THE web language

Unix predated it
Unix became THE web server

Who had trouble?

Microsoft w/Win 95 and 98
And Windows 95 came *after* Mosaic

Controlling change is death

or at least really painful in Microsoft's case

90% is crap

Taking a big crap is healthy

One man's crap
Is another man's gold

App::Asciio - Plain ASCII diagram

| | | |
| | | | | |
| | | | | |
v | v | v |
v v v
______ _____
/ _ __ __/ __`
L ____ ___ /_/_ /
-----> __ /',__ /'___/ / ----->
/ /__, `/ __/ _
_ _/____/ ____ _ _ _____
/_//_//___/ /____/ /_//_//_____/

| | | |
| | | | | | |
v | | | v | |
| v | | |
v | | v
v v
(_/)
(O.o) ASCII world domination is near!
(> <)

default elements in an empty diagram
.------------------------.
| | |
| | |
.-------|---------------|--------|------------------------------------------.
| | | | asciio |
|-------|---------------|--------|------------------------------------------|
| ......v---------......v........v......................................... |
| .....| |..edit_me....--->....................................... |
| .....'----------'........................................................ |
| ......................................................................... |
| ...................----------------..---------------..................... |
| ..................| ASCII > || thin_box |.................... |
| ..................| Rules line > || text |.................... |
| ..................| Load || wirl_arrow |.................... |
grid------------->..........| Save || arrow |.----------......... |
| ..................| Export > || box > || star_box |........ |
| ..................'----------------'| Rulers > |'----------'........ |
| ...........................^........| Misc > |.................... |
| ...........................|........| T_star |.................... |
| ...........................|........'---------------'.................... |
| ...........................|............................................. |
| ...........................|............................................. |
| ...........................|............................................. |
| ...........................|............................................. |
| ...........................|............................................. |
| ...........................|............................................. |
'----------------------------|----------------------------------------------'
|
context menu

.---. .---. .---. .---. .---. .---.
OS API '---' '---' '---' '---' '---' '---'
| | | | | |
v v | v | v
.------------. | .-----------. | .-----.
| Filesystem | | | Scheduler | | | MMU |
'------------' | '-----------' | '-----'
| | | |
v | | v
.----. | | .---------.
| IO |<----' | | Network |
'----' | '---------'
| | |
v v v
.---------------------------------------.
| HAL |
'---------------------------------------'

???? ??? ????? ??????
(Nadim Khemir)

He had no idea if this would be a cool thing

He's written all this other stuff that hasn't gotten a lot of attention
He happened to mention asciio at the hackathon
Everyone made them show it off
It was cheered, surprise hit of the conference

Will this be useful?

People spend a tremendous amount of time worrying if something will be useful to anyone
else.
Should they release it?

No release
==
No information

You can't predict use
Unless you release it you'll never know if anyone's going to use it.

Don't worry,
Be crappy

Guy Kawaski
Just throw it out there and see what sticks.

Release Early,
Release Often

Because you don't know what's going to be crap and what's gold
Throw it out there and ?nd out
If nobody uses it, no worries, work on something else
If the ?rst release is crap and it gets used, you can always ?x it in the next one

Cheap releases

The release process should be so cheap that the cost of release is not a consideration
I automate most of the process

Low barriers to entry

CPAN only requires your release is...
freely re-distributable (not even Open Source)
something to do with Perl

Feedback about use

The author should get lots of feedback about their release, if it's getting used, if there's bugs,
patches, critiques...

Feedback Loop

Starts a feedback loop
Author makes changes based on use

Evolution

System evolves based on use
Evolution is very healthy and very robust

How do you know what
people need?

Nadim came up with this great thing, asciio
Why did it he do it?

You know your needs.

Instead of writing what others need...

Write what you need.

Knowing and doing what you need for your situation is known as...

Necessary

It is the knowledge you need.
It solves the problems you have.
It includes all the necessary hacks to make it work in the real world.

As opposed to...

About control

Top down control.
Making it easier for the implementers at the expense of the users.

Finding the holy grail
Often never found

About getting things done

Metis is...

You know best if the
solution ?ts

quot;Art of muddling throughquot;

You write what you need.

Then release it and see if anyone else needs it.
Astoundingly effective.

Look side to side

Look to yourself and the people around you
They have a similar problem
They'll accept a similar solution

Even the best authors only make a handful of useful modules

Write for yourself

It's ok to be sel?sh
The primary reason for working on something is because you need it
Not because others need it
Working for others for free means burnout.

Work for yourself
Share with others

�ݺ�ߣ

Crap Is Gold

More Related Content

Crap Is Gold