Chapter 5. Repositories

The term "On-Net" or "On-Disk" are used here interchangeably to represent an operating-system level repository that consists of a number of files that might be accessed by some physical storage mechanism such as a USB stick or might be accessed over a network interface using the HTTP protocol for example.

Repositories are a collection of packages together with some meta-data or index about the packages. A Haiku-OS concept of a repository is a set of URL-accessible resources. A base URL might be something like;

http://packages.haiku-os.org/haikuports/master/repo/x86_gcc2/current

At this location, a number of resources should exist that have well-known names so that they can be easily found. The resources in a repository are generally targetting a specific architecture such as "x86_gcc2", but there may also be resources for "any" architecture as well.

name HaikuPorts
vendor "Haiku Project"
summary "The HaikuPorts repository (for Haiku R1 BETA3)"
priority 1
identifier tag:haikuports.org,2013:repositories/haikuports/r1beta3/x86_gcc2
architecture x86_gcc2
                

Haiku Depot Server's (HDS) concept of a Repository is, in nomenclature, slightly mis-aligned with the On-Net concept. An HDS Repository represents the collection of On-Net Repositories that are conceptually serving the same or similar packages for different architectures. A RepositorySource is a placeholder for the different architectures. A RepositorySource may then have one or more RepositorySourceMirrors in order to store a URL to the repository data for a given geo-location identified by a country code; for example...

The following diagram illustrates the relationship between the Haiku Depot Server Repositories and the On-Net repositories of the operating system. The yellow elements are the HDS elements and the blue elements are the on-net elements.

Figure 5.1. HDS Repositories and how they relate to On-Net Repositories.

In this artificial example, an HDS Repository "Peanut" exists. Consider for example that a given system is x86_gcc2 architecture and is located in New Zealand then it would be possible to follow the path through the Repository to the RepositorySource to the RepositorySourceMirror in order to obtain the Base URL to be used in order to access the On-Net files of the repository.

HDS has its own view of an 'on-net' repository and it is possible that the 'on-net' repository has been updated and so is newer than the HDS view of the same data. This section discusses how the HDS data can be updated so that its data is synchronized with the remote 'on-net' data.

Ultimately what happens is that through some mechanism, HDS is prompted to fetch the remote data and to import it. A periodic process will run that will cause HDS to run from time to time and to import the remote data for each RepositorySource. In a specific example, for "haikuports" repository for x86_64, each day, HDS will pull down the HPKR data and will import it.

Often the 'on-net' remote repository data is updated. This might happen because a package has been updated and the remote HPKR data is re-built with the new package data included in it. At this moment, after the build is complete, it would be ideal for HDS to be prompted to poll the repository in order to pull down the HPKR data and update its own records. HDS provides a simple API for this and the administrators of repositories are invited to use this API in order to cause HDS to update its own records in a timely fashion.