OK, so that's a weird behavior :) I'm the author of the aoepeople/composer-installers package and the idea behind this is that this package provides an alternative (very basic and simplified implementation of the composer package-type magento-module
- and adds another type magento-source
. I only want the installer to put the package in the right place - keeping it simple. That's why I decided to come up with an alternative package.
The reason why the aoepeople/composer-installer replaces the magento-hackathon/magento-composer-installer is because many Magento modules already come with a composer.json that requires the magento-hackathon/magento-composer-installer. In order to be able to seamlessly use the simpler installer that one is "pretending" to be the hackathon-installer. But unless you actively require aoepeople/composer-installers in your project's composer.json you should continue using the original installer as no Magento module's composer.json out there (not even ours) is directly referring to aoepeople/composer-installers.
The fact that packagist tries to be "smart" and returns a package that replaces a package that's not registered is new to me and - to be honest - very disturbing. While this might be intended behavior I share the opinion that this could easily be abused. I start liking Alan's idea to bypass packagist completely using "packagist":"disabled"
. Especially in the case of Magento modules this seems to happen easily because most of the Magento modules are unknown to packagist while being registered at packages.firegento.com only.
A simple quickfix/workaround to this specific situation could be to register magento-hackathon/magento-composer-installer on packagist.org.