CPAN 6 Bottom — UP German Perl Workshop

CPAN 6 Bottom - UP German Perl Workshop, 21 feb 2007 Mark Overmeer

CPAN is. . “CPAN is the best thing that ever happened to Perl” GDB#1

CPAN is. . “CPAN is the best thing that ever happened to Perl”

CPAN is. . “CPAN is the best thing that ever happened to Perl 5”

CPAN is. . . Just over 10 years younger than Perl 5 WWW UNIX

CPAN is. . . Just over 10 years younger than Perl 5 WWW UNIX . . . developed before break-through of XML/SOAP/WSDL CVS/SVN/SVK/GIT spam, viruses unicode

. . . concerns. . . 11 k distributions, a lot is bad quality or poorly maintained 5700 authors, who is who? who is making the rules of CPAN?

. . . concerns. . . 11 k modules, a lot is bad quality or poorly maintained 5700 authors, who is who? who is making the rules of CPAN? GBD#2 What will happen when infected code is detected on CPAN? Will Perl loose its best thing?

. . . future. . . Where will we store Perl 6 modules pir, pbc, pod only, let the big name-space battles commence! Par. Tcl? Brainf*ck? etc

C Perl 6 Pod Pir Pie. Ton Par. Tcl Pbc 6 AN = CP AN 6 CPAN 6

The Beginning

The process Autho r CPAN modules Users

The process Autho r Read ers Store Books

Release Why do you release material? to share with many people

Release Why do you release material? to share with many people When do you release material? when it is ready

Release Why do you release material? When do you release material? to share with many people when it is ready What can be “released”? coherent collections of bytes in files

Release Why do you release material? When do you release material? when it is ready What can be “released”? to share with many people coherent collections of bytes in files How do we release materials? ftp-servers (CPAN), CDs/DVDs,

Releases ≠ Revisions a logical sequence of development steps of a file each step may solve some bugs each step extends the code and documentation steps are created by a single person Releases are created to be distributed a defined `frozen' state of the software (some combination of file versions) published by someone who is responsible

CPAN ≠ YA VCS CPAN is a collection of releases distributing information name-spaces and trust searching in the collection Version Control Systems manage (related) (development) revisions of files collecting pieces of information traceable changes tracing and solving bugs selected revisions of selected file are released

What is interesting? publish the information authors, publisher

What is interesting? publish the information authors, publisher storing the information stored where for how long who sets the rules

What is interesting? publish the information storing the information authors, publisher stored where for how long who sets the rules download the information search, distinguish who can download: user

What is interesting? publish the information signatures storing the information authors, publisher stored where for how long who sets the rules archive rules trust download the information search, distinguish who can download name-spaces access

What is interesting? publish the information signatures pause-id archive rules trust ftp. cpan. org ever Andreas & co storing the information authors, publisher [CPAN] stored where for how long who sets the rules download the information search, distinguish who can download name-spaces one access everyone

Out of scope for the end-users: how the release was created, as long as it is “stable” and “trustable” for the authors: what the end-user is doing with it, as long as he/she has access permissions.

Out of scope for the end-users: how the release was created, as long as it is “stable” and “trustable” packaging tools: Maker/Module: : Build development versions: CVS, . . . for the authors: what the end-user is doing with it, as long as he/she has access permissions. installation tools: CPAN. pm/CPANPLUS

Out of scope for the end-users: how the release was created, as long as it is “stable” and “trustable” packaging tools: Maker/Module: : Build development versions: CVS, . . . for the authors: what the end-user is doing with it, as long as he/she has access permissions. installation tools: CPAN. pm/CPANPLUS Data distribution can be content independent

Really unimportant how the data it is packaged tar, zip, separate files

Really unimportant how the data it is packaged tar, zip, separate files how it is compressed gzip, bzip 2, diff

Really unimportant how the data it is packaged how it is compressed tar, zip, separate files gzip, bzip 2, diff how it is transported ftp, http, mail, tape, printed

Really unimportant how the data it is packaged how it is compressed gzip, bzip 2, diff how it is transported tar, zip, separate files ftp, http, mail, tape, printed where it is actually located ftp mirror, cd/dvd distribution, git depot

Really unimportant how the data it is packaged how it is compressed gzip, bzip 2, diff how it is transported tar, zip, separate files ftp, http, mail, tape, printed These could best all be hidden to publishers and end-users where it is actually located ftp mirror, cd/dvd distribution, git depot

Needs for Perl 6 place for Perl 6, Pir, Pasm, Par. Tcl, . . . requires Perl 5 to install Perl 6? extended versioning scheme explicit version parallel authoring no deletions: huge pkg list DBI-TIMB, DBI-MARKOV

Problems with Perl 5 Interface breaking changes spamassassin 2 vs. spamassassin 3 Version of distribution/which distribution contains installed package. Long @INC search paths for pm files Impossible to oversee effect of an install request to the system

NO QUICK FIX! but a good alternative

CPAN 6 because Compared to CPAN a wish for information independent: reuse of coding effort more name-spaces: DBI for perl 5, DBI for perl 6 more security; PGP/SSL authentication clear maintenance rules; no GBD archive “tricks” like subsets (cpan-mini) remove ad-hoc configuration of mirrors complex search, server-side know before download: license, nested dependencies, . . . release deprecation, release under embargo and more

CPAN 6 definitions Release: a set of files distributed for a specific purpose to end-users. Project: a set of releases, sharing the name but with different version identifiers. in Perl 5: “a distribution” a name-space entry Archive: a set of projects, collected for some purpose. storage, rules, entry-point an URI

Collections Perl 5 distributions Perl 6 (& friends) distributions conference materials wikipedia articles linux distributions photo or music collections company's press releases newspaper articles cooking recipes pgp public keys use case

Split CPAN 6: idea of having a collection of releases, combined into archives uploading releases downloading releases distributing the archives combining archives filtering archives like mirrors like ftp sites like cpan-mini CPAN 6 is the DISTRIBUTING

Split CPAN 6 is the distributing (network) Pause 6: handles one archive is one name-space is set of releases with some functional relation a set of releases relate to a project(name) projects are allocated to authors have a well-defined identity an archive has well-defined rules, by a board Pause 6 is ADMINISTRATION

Split CPAN 6 is the distributing (network) Pause 6 is the administration (archives) CPAN 6. pm installs Perl 5, Perl 6 etc distributions blends into the platform collects meta-data to publish releases programming language specific component CPAN 6. pm for INSTALLATION

Split CPAN 6 is the distributing Pause 6 is the administration CPAN 6. pm installs releases locally Of course perl -MCPAN 6 'install Test: : More' will work!

Pause 6 is just one possible implementation of the CPAN 6 ideas Releases have project-name utf 8 version utf 8 state type loads of (optional) meta-data follow-up information

Release strands 0 = allocate project name and get name-space rules solid arrow: non-breaking follow-up dashed arrow: code fork

Release states Release state changes require signatures Release states are uploading, initiated by publisher published, distributable to deployers embargo, uploaded but blocked released, by the authors deprecated, by the authors expired, by the authors rejected, by the board/authors installed, by the end-user

Project types project types are: publication constitution identity archive reference license description by user, set of files archive configuration public keys, . . . therefore everything follows same release protocol one transport implementation covers all aspects

Use Cases Board writes constitution Changes in constitution Opening and restricting name-spaces Authors start project Publisher uploads a release

CPAN

CPAN 6/Pause 6 terms Commissioner search for end-users Stores point to upload to, for publishers executes constitution for the archive board Deployers CPAN Pause download locations Scribes CPAN mirrors transport processes publishers to commissioner, comm. to deployers, deployers to users, archive hierarchies

CPAN 6 concept

Pause 6 Implementation

Hierarchy Scribes can be configured to create archive hierarchies, for example: global cpan-perl 5 business level sub-set department sub-set system local personal installed personal development

CPAN 6 network

Use Case Ftp-server with mirrors: initializing a deployer User contacting new archive

Status CPAN 6/Pause 6 global design document CPAN 6 design and implementation doc Pause 6 design and implementation doc Start of an implementation: the creation of large building bricks

Promises All data handling UTF 8 All data definitions in XML schema's No new data exchange protocols use HTTP, FTP, RSYNC, . . . No new communication mechanisms use SOAP, RPC, . . . No new security systems use PGP, SSL, MD 5, SHA-1, . . .

Coherent Documentation Major parts of documentation are generated. Saved Mail: : Box 700 k characters OODoc A must for OO programs exist for many years, used for many years

Data Exchange I hate XML, but. . .

Data Exchange I hate XML, but. . . I hate SOAP: : Lite, so. . .

Data Exchange I hate XML, but. . . I hate SOAP: : Lite, so. . . XML: : Compile in usable state, for this CPAN 6 purpose

XML: : Compile Under control of a schema pure Perl real XML real XML pure Perl perl template (example) XML template (example) Validating, name-spaces, any/any. Attribute, substitution. Groups, WSDL, SOAP, fast No XML in your Perl program!

Platform independence OS independence File: : Spec File: : Basename File: : Copy File: : Remove File: : Temp File: : Path File: : Find often used, OS related Cwd Slurp tainting delete $ENV{PATH} $^O IO: : File: : stat Sys: : Hostname $0 $PROGRAM

Platform independence No File: : Spec, File: : Basename, File: : Find, etc Path: : Class: my $dir = dir 'foo', 'bar'; # Path: : Class: : Dir my $file = file 'bob', 'file. txt'; # Path: : Class: : File # 'bob/file. txt' on Unix, 'bobfile. txt' on Windows print "file: $filen"; my $subdir = $dir->subdir('baz'); # foo/bar/baz my $parent = $subdir->parent; # foo/bar my $parent 2 = $parent->parent; # foo my $dir 2 my $text = $file->dir; = $file->slurp; # dirname = bob

Platform Independence No File: : Spec, File: : Basename, . . . Path: : Class super-set: No Cwd, $^O, delete $ENV{PATH}, etc Sys: : Indep first steps made

Sys: : Indep use Sys: : Indep devel => 'linux'; my $dir = dir '/etc/'; my $f 1 = file '/etc/passwd'; # abs my $f 2 = file 'here. txt'; # rel my $cwd = program->cwd; my $f 3 = file os->home, '. ssh', 'known_keys'; print os->family; # 'unix' my @m = os->volumes; # mount-points # filename mutilated based on file-system rules my $fh = $f 3 ->open('r'); program->find(. . . ); OS Program File/Dir Filesys

User Interaction Would like to have command-line for scripting (many options) interactive commands, like IOS Curses: : UI Tk / wx. Widgets Gtk / Qt XHTML / XUL Transparently based on GNU gettext() schema generated components set of standard code building bricks ?

where we are. . . . discussion and attention contacts with other potential user groups signing, licensing, trust algorithms still a lot of detailed design to do funding to create base implementation money, man-power, test setups 30 k€