| Module Info
| Add a review of Net-FullAuto
This is a response from the author to the two reviews posted below. For a very long time, Net::FullAuto was experimental software, and for years there were warnings in the sparse documentation that this software was not ready for general use. That said, at the time the reviews below were posted, many of the points raised in these reviews were genuine issues, most of which have since been remediated or explained. The POD documentation is now complete and up to date. It bears little resemblance to the sparse documentation these reviewers encountered more than a year ago. That said, the accusation of malicious intent without and any evidence to support this assertion other than complexity the reviewer had no wish to navigate, is irresponsible - and that is being kind. His very words "I suspect" demonstrate his lack of true inquiry, and his eagerness to denigrate and even slander the author without adequate foundation, is something the author hopes is transparent to others. That said, I will address these concerns because some of them are valid, and are concerns that others will share.
Again, documentation is NOW complete, up to date, and accurate.
"Marketing" style language has been mostly removed, and any that remains is informational and more of what one would expect to find in "documentation".
It's true that the installation of this module is highly customized. Net::FullAuto has over 100 dependencies on other modules, and many of those modules have long chains of their own dependencies. To get all these to install successfully in a 100% automated fashion literally took years to achieve. Conventional Perl distribution tools were largely abandoned because they simply couldn't do the job without requiring significant manual intervention and expertise from the user. Many of these tools have since evolved, but there are still multiple reasons that they cannot yet provide a complete solution for installing Net::FullAuto. Net::FullAuto is workload automation software designed to compete with packages like Chef, Ansible, Puppet and others. Anyone familiar with those frameworks know how complex and challenging they are to install and work with. The goal of the FullAuto project was to come up with a framework that was not only easier to work with than these others, but significantly easier. That goal is met only with a fully automated installation of the tool and all its dependencies. It's hard to promote "workload automation software" as ground breaking if it fails to fully automate its first genuine workload - its own installation.
It's true that the CPAN TESTERS service has been disabled. Because of the large number of dependencies, The Net::FullAuto install uses "continuous integration". In other words, dependencies will always be updated to the latest version. The CPAN Tester community use hosts with very sensitive configurations, and encountered problems with modules being updated automatically. To conform to the needs of the CPAN Testers would mean putting an incredible burden on ordinary users requiring them to perform dozens (even hundreds) of manual installation tasks they have neither the time, patience nor expertise for. Users are warned of this behavior when first endeavoring to install Net::FullAuto, and are advised to use a host, or alternate secondary Perl installation where automated updates of modules is not an issue.
'sudo' is used only with Amazon EC2 hosts as of this writing. Net::FullAuto needs administrative privileges to be installed on EC2 hosts, and users have to go through elaborate setup steps to enable an Amazon account with the necessary privileges. Their consent is assumed for these reasons.
There are a number of small utilities bundled with the distribution. They are all open source and free of any restrictions. They may be removed and stored elsewhere (like sourceforge or github) in the future. The use of Figlet fonts is explained in the POD documentation.
It is true that a number of accepted CPAN standards have been abandoned or significantly altered. There are good reasons the standards exist, but also good reasons for having to avoid them. The decision was made that a FULLY automated installation of Net::FullAuto and ALL of its hundreds of dependencies justified the decisions made. All are subject to change as both Net::FullAuto and the CPAN evolve.
Net::FullAuto is unique as it uses a PERSISTENT connection to communicate with remote nodes. This functionality took 16 long years to perfect, but is now ready to do the kind of robust workload automation only dreamed about - until now.
I hope you will give it a chance to impress you, and I hope you will choose to open tickets and work with me and my team to improve Net::FullAuto, rather than post a permanent negative review over a temporary concern.
If safety and security are concerns - as they should be, please experiment with Net::FullAuto on a public cloud or other disposable machine image before installing on a more sensitive host. This video will walk you through the entire setup process on Amazon EC2:
This module should be avoided by all. Personally, I suspect its use will cause damage either due to author's incompetence or malicious intent.
The pod reads like a bad Infomercial: "IT REALLY IS THAT EASY!", "Net::FullAuto is POWERFUL", "Net::FullAuto is the ANSWER!"
It barely describes what the module does, and according to the extensive RT#100658 (titled "documentation lies") the module performs many security-sensitive tasks, yet they are undocumented and the unmaintainable mess of the code essentially guarantees there are droves of bugs. Furthermore, the output (according to the docs) of the script that uses this module is litered with unwanted messages like "Starting fullauto" and asking the user for additional passwords, which limits the use of this code as a Perl module.
The author doesn't use conventional Perl distribution tools and seems to be manually hacking Makefile.PL, which, at the time of this writing, is 6319 lines long. The code is trying to guess OS in use and then installs a bunch of software using system package manager (e.g. sudo yum -y install 'openssl-devel'). Due to issues, the author also disabled CPAN TESTERS service on the module.
The installation asks for sudo password for no explained reason (RT#97421).
Looking at the files inside the distro, there are lots of puzzling examples: UNINSTALL_CYGWIN, install_mozrepl_plugin.ahk, a directory with 75 figlet fonts! What are those doing in a Perl CPAN distribution? The code is a huge mess and even the author says it isn't really meant to be understood by merely perusing it (RT#100658).
The author also frequently uploads null updates to CPAN, presumably to gain exposure to the module, thanks to CPAN update services. The current version number is a ridiculously long sequence of 9s and the 7-year old Changelog contains 2278 lines!
In summation, this distribution does not adhere to accepted CPAN standards, the code is in poor state and works with security/system sensitive aspects, which makes the code dangerous to use.
Avoid at all costs. Don't even install for curiosity.
Continuously spams CPAN update services like the IRC bots and RSS feeds with multiple null updates every day. All functionality is in one giant script. Rambling documentation full of SHOUTY ALL CAPS sections that conforms to no comprehensible format. Recommends installing itself as setuid.
Not to put too fine a point on it, but don't even consider touching this with a bargepole.