Notice: This Wiki is now read only and edits are no longer possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.
Equinox p2 1.0 Technical Specs
Contents
User Interaction
- Managing other profiles
- p2 inherently supports the management of profiles other than the one currently running the p2 infrastructure. Priority: 1
- * Mechanism to identify the profile managing a product
- * Mechanism to identify the agent (agent data area) managing a product
- * UI to browse profiles. This may be used in the target management as part of the tooling.
- [john - What we have today is p1, but anything more fancy is p2]
- Drag and drop installation
- Users can install new function by dragging and dropping the related files (JARs, directories, zips, ...) on a running Eclipse. Priority: 2
- * Relationship between this and the directory watcher
- * UI to support this kind of metaphor
- Browser-based installation
- Installation of new function can be triggered by users clicking on a link in a browser. Priority: 3
- * Mechanism to find running instances of eclipse
- * Mechanism to invoke a p2
- Directory monitoring
- Users can designate any number of directories to be watched. When installable elements (e.g., bundle JARs and directories) are copied into or remove from watched directories, their contents are installed or uninstalled (respectively). Priority: 1
- * Support for installing metadata-less bundles (Need to invoke the metadata generator)
- * Support for installing metadata-aware bundles. Need to define a serialized format of IUs for embedding with the bundle. Do we have a concept of artifact repository.
- * Mechanism to control the frequency at which directories are being polled
- * UI to add/remove watched folders and set the polling frequency
- * Define backward compatibility with UM links and extensions folders
- * Integration into the startup sequence.
- * Relationship with the shared install and the reconciler
- * Support for multiple bundle pools
- [john - some of these things should be p2. Specifically, we need to support adding extra folders, and some way to kick off a reconcile, but dynamic behaviour is p2]
- Update scheduling and policies
- p2 supports a number of update policies allowing users to, for example, set update polling periods, schedule updates and have detected updates automatically downloaded. Priority: 1
- * See also Staged provisioning.
- * Need a way to notify the user about available installations. Priority 1.
- * Need a way to set the frequency of checks. Priority 1.
- Headless operation
- All p2 function is accessible through command-line or programmatic interfaces. Complete installation operations can be performed without a graphical user interface. Some operations support the use of response files to silently provide input. Priority: 2
- * Support to get the input from a response file. Define this in relationship with the callable User Input service.
- * Define applications to invoke p2 from the command line to perform installation / uninstallation of IUs. See also http://help.eclipse.org/help33/index.jsp?topic=/org.eclipse.platform.doc.isv/reference/misc/update_standalone.html
- * Separation of the installation phase from the configuration phase. Priority 1.
- Remembered signers
- Users are able to accept signer certificates and not be asked each time such remembered certificates are encountered. The set of remembered certificates is managed through a preference page that allows for certificates to be added and removed. Priority: 2
- * Storage for remembered signers
- * Mechanism to present signatures, see also callable User Input.
- Integrated user prompting
- Prompts for security information (e.g., login, certificate trust, ...) are consistent and well-integrated into the user workflow. Priority: 2
- * Callable User Input service to gather information from the User (https://bugs.eclipse.org/bugs/show_bug.cgi?id=206903). Priority: 2
- * Presentation of licenses. Priority 1.
- Internationalization
- p2 will provide an internationalized installation experience. Priority 1.
- * Support for DBCS in the metadata, the artifact names, the install paths. Priority 1.
- * Ability to install a translated version of a software. Priority 1.
- * Ability to translate the content of the metadata while browsing the repository. Priority 2.
Download technology
- Automatic detection of proxy/socks settings from the OS/Browser
- p2 uses the standard Eclipse proxy and socks settings management system and integrates such settings from the OS and browsers. Priority: 2
- * (ECF) Ensure platform proxy settings are being used. Priority 1
- * (Platform) Detection of OS / Browser proxy settings
- Adaptive downloads and mirror selection
- p2 dynamically adapts its artifact download strategy based the characteristics of the servers available, the connection speeds and the system being provisioned. Retries are automatically attempted and mirrors re-selected depending on failures. Priority: 1
- * Support to get stats as we go
- * Support to query repo for partially obtained artifacts
- * Interaction of restartable download with processing steps
- * Query property from artifact repositories
- * Query transports available
- [john - we should refine priorities here. We need mirror selection, but how fancy does it need to be? I.e, perhaps we don't need adaptive mirror strategy within the download of a single artifact. A simple strategy would be to pick a random mirror for each artifact, and discard mirrors that are unresponsive or that fail. In case of a mirror failure the partial artifact could be discarded.]
- Download integrity through MD5/SHA1 and signature verification
- The integrity of downloaded artifacts can be verified using MD5/SHA1 hashing algorithms and/or signature verification. Priority: 1
- * SHA1 processing step and publication steps. Priority 2.
- * MD5 processing step and publication steps. Priority 1.
- * Signature verification processing step. Priority 1.
- [john - I suggest MD5 is p2, since signature verification achieves a similar effect]
- Integrated compression technologies
- p2 allows artifact repositories to maintain artifacts in a variety of formats (e.g., compressed using pack200, JAR and binary deltas relative to previous versions, etc.). This can dramatically reduce the bandwidth requirements for new software installations. Priority: 1
- * Pack 200 processing step and publication step. Priority 1.
- * General delta processing step and publication step. Priority 2.
- * Jar delta processing step and publication step. Priority 2.
- * See also repository optimizers tool.
- Peer-to-peer downloads
- Since all downloads in p2 are based on a mirroring metaphor, artifacts and metadata can come from repositories on central servers or peer machines on a local network. Priority 2
- * Publication of local repositories as http. Priority 2.
- * Announcement of available repositories on the network (e.g. SLP). Priority 2.
- Transparent restart
- Aborted installations and downloads can be restarted without refetching. Priority: 1
- * Artifact repositories as pools. Priority: 1
- * Garbage collector of artifacts. Priority: 1
- * Repositories to support restart for failed downloads. Priority 2.
- Download time estimation
- Estimation of the download time as the download progresses. Priority: 2
- Repository seeding
- Repositories are able to reference other repositories and thus inform p2 of additional sources of artifacts and metadata. Priority: 1
- * Define a way to discover other repositories and add them. Priority 1.
- Media support
- p2 supports and properly manages the interaction with repositories stored on removable and volume-oriented media such as CDs, DVDs. Priority: 1
- * Define repository format for volume-oriented repositories. Priority 1.
- * Ensure compatibility of the smart download manager with such medias. Priority 1.
- [john - is this really p1? Did UM support this?]
Security
- Metadata signing
- Metadata is signed to ensure content integrity. Priority: 3
- Secure transports (https, ...)
- Secure transports such as https are supported. Priority: 1
- * (Security) Certificate storage
- * (ECF) Support for https connection
- * (ECF/Security) ECF / Security integration for prompting and storing certificates.
- Repository trust
- p2 has the ability to identify repositories as trusted or untrusted as well as white and black lists of domains housing repositories. Priority: 2
- * Define white list / black list and integrate into the repository managers.
- * Define a notion of trusted repository. How is the trust established. What are the implications of being trusted. Where is this information stored.
- Repository authentication
- p2 supports a variety of mechanisms for authenticating to servers. Priority: 2
- * Support to query the user for login / pwd. Priority 1.
- JAR signature verification
- Signed JARs downloaded from untrusted repositories are verified to establish trust. Priority: 1
Core facilities
- Generic Metadata
- Underlying p2 is a generic metadata model of Installable Units. p2 metadata captures dependencies on non-Eclipse/OSGi based elements (e.g., JREs, native code, other applications, ...) as well as on physical elements of the machine (e.g., number of CPUs, amount of memory or drive space). Priority: 1
- * Define a generic model of capabilities / requirements. Priority 1.
- * Define the way by which some capabilities get populated. For example how do we know the OS. See also JRE reconcialiation pb which is similar. Priority 1.
- * Separate meta-level capabilities and requirements form the base level ones.
- * Separate the dependency information from the configuration one. Priority 1.
- * Support for canonicalization of the version numbers. Priority 3.
- * See also Fix delivery
- * See also Update delivery
- Shared (multi-user) installs
- Scenarios where Eclipse installs are shared across multiple users is streamlined. Priority: 1
- Bundle pooling
- p2 pools the set of bundles installed across the profiles it manages such that any given bundle appears only once on disk. This saves disk space as well as dramatically speeding subsequent installation operations. Priority: 1
- * Support for multiple bundle pools per profile. Priority 1.
- Garbage collection of unused bundles
- Bundles no longer used in any managed profiles are garbage collected according to a flexible policy. Priority: 1
- * Bundle pool garbage collector. Priority 1.
- * Metadata garbage collector. Priority 2.
- * Need a way to ensure that a bundle pool is only used by one agent to avoid problems with GC. Priority 2.
- Resilience to install problems
- p2 provides a best effort approach to ensure that failed installations do not leave the system in an inconsistent state. This includes a safe mode for the provisioning infrastructure itself. Priority: 1
- * Define a way to ship an immutable agent installation within an eclipse install. Priority 1. [john - is this referring to the admin UI? Or is this just a copy of "bundles.txt" stored in a safe place to use for failure recovery? To me, "resilience to failure" is the p1 item, this item is just one possible implementation approach.]
- * Resilience to corrupted files. Priority 1.
- * The engine needs to persist the operation currently being performed. This allows for reboot of the provisioned system during an installation, as well as for checkpointing on restart in case the system crashed during an installation. Priority 2.
- * Leave the user machine in a clean state when the installation is cancelled. Priority 1.
- ** Touchpoint actions need to capture previous state before doing their work. Priority 1.
- ** Support for Touchpoint actions to store previous state. This allows for recovering from a crash. Priority 1.
- Fix delivery
- Fixes to existing installed function can be installed and uninstalled without updating the base function. Priority: 1
- * Define metadata for fixes. Priority 1.
- * Define the lifecycle of the fixes, when do they get installed, how do they get discovered, when are they uninstalled. Priority 1.
- Sequenced provisioning
- Users and developers can mandate that various update and install operations must be executed prior to attempting subsequent operations. Priority: 2
- * Define markup identifying such updates.
- * Adapt the director to detect such updates and process them accordingly.
- Staged provisioning
- Provisioning operations can be staged such that all required artifacts are downloaded and then, at some later time, the actual installation and configuration executed. Priority: 2
- * Ability to invoke the engine by specifying different phases.
- Fine grain installation
- p2 supports the installation of individual Installable Units as well as groups of Installable Units. Since typically one IU represents one bundle, p2 allows for the installation of individual bundles. Priority: 1
- Dynamic dependency discovery
- When p2 is asked to install an IU it can optionally attempt to satisfy all prerequisites by discovering and installing other IUs that supply the required capabilities. Priority: 2
- * Explore a resolution algorithm supporting backtracking. Priorty 1.
- Managing non-running systems
- p2 is able to manage Eclipse profiles even when the profile is not active/running. Priority: 1
- [john - this needs more detail. Does it mean the user can do this from a GUI?]
- Managing running systems
- p2 is able to manage and properly interact with running Eclipse profiles. For example, triggering restarts of the remove system as needed. Priority: 2
- * Ability for touchpoint to cause a runtime to be started / stopped / restarted. Priority 2.
- * Ability to update eclipse.exe
- Rollback
- Users can restore previous states of a profile. Priority: 1
- * Remember previous state of the system at the profile level and allow for returning to a previous state. Priority 1.
- ** Need to name previous state and identify the cause of the changes. Priority 1.
- ** Need to group/batch the changes done on a profile and have an API allowing for such a thing. Priority 1.
- ** Decide whether or not this is core to the agent. Priority 1.
- ** Define a mechanism allowing to set the purge policy. Priority 1.
- Profile interchange
- Profiles can be manipulated and exchanged between users. This allows previous setups to be stored and recreated and for users to exchange profiles. Priority: 3
- Revert to the previous install
- When an installation succeeds but is not satisfactory, users can revert the system to the exact same state as it was before. Priority: 3
- * Remember previous state of the machine at the system level (e.g. snapshot the file system, the registry, etc.). Priority 3.
- ** Define a phase and an API allowing for such operation to be done. Priority 3.
- ** Provide a preference to disable such a mechanism and to also purge.
- OS integration
- Applications installed using p2 can be tightly integrated with the underlying operating system. For example, desktop shortcuts, registry entries, etc. can be deployed as part of the installation. Priority: 2
- Governor
- The governor represents an authority allowing or vetoing provisioning operations. Priority 2.
- * Review if the AccessController like model is suitable.
- * Define which operations must be protected.
- * Define how it gets the context in which the operation is executed.
- Framework Admin
- Framework admin is an API allowing for the transparent manipulation of OSGi frameworks. Priority 1.
- * Review performance on large systems.
- [john - is all of framework admin p1, or just ability to modify eclipse.ini/config.ini?
- Simple configurator
- Simple configurator is the bundle responsible for taking the description of a system and applying it.
- * Validate performances on large configurations. Priority 1.
- * Provide an option to have simple configurator play nice with others (don't delete bundles it does not know about). Priority 1.
- * Decide if simple configurator checks for the start levels programmatically. Priority 1.
- * Review the way start levels are being changed. Priority 2.
- Misc.
- * Nested profiles. Ability to have profiles refer to each others. Priority 2.
- * Variables. Ability to parameterize the installation and configuration of an IU. Priority 2.
- * File locator. Ability to locate a file within an IU (Priority 1) and from another IU (Priority 2). This has an implication on the actions who refer to files and who lay down files.
- * Agent data sharing. Need a way to share the agent data across multiple running processes. Priority 1.
- * Flavors. Review the concept of flavor and the default flavor we ship. Priority 1.
- * Markers. Review and refine the concept of installable unit markers and their lifecycle. Priority 1.
- * Make the agent dynamic. Priority 2.
- * Scalability. Priority 1.
- * Reconciliation with some "changeable" part of a profile (e.g. JRE).
- [john - I think only scalability from above list is p1. We don't have time to focus on these conceptual things - should only work on them if they directly help with some other p1 item]
UM Compatibility
- Update site integration
- See also Update Manager and p2 Compatibility
- p2 is able to read existing update sites created for use with Update Manager. Indexing and conversion tools are provided for optimizing the use of such sites. Priority: 1
- Metadata generator tool to run over update sites. Priority 1.
- Support for mapping update site categories in p2 style categorization. Priority 2.
- Define representation of categories in p2. Priority 1.
- Support to install from pure old-style update site. Several alternatives:
- Use the update manager code and have it generates the bundles.txt. We may still use the new UI but in degraded mode.
- On the fly generation of Metadata / artifact repos, reusing MayInstall code.
- Feature compatibility
- The feature data structure is not part of the p2 infrastructure but p2 allows Update Manager Features to be published to metadata and artifact repositories. Priority: 1
- Does p2 continue still deliver the features out of the box?
- When a user uses 3.4 and it has features depending on the features from the SDK, what happens?
- Feature Install Handlers
- Feature install handlers continue to work in p2 with some restrictions. A migration/porting guide helps developers in moving to the new infrastructure. Priority: 1
- Links directory
- p2 includes tooling to publish existing Eclipse installs into metadata and artifact repositories. This :includes the correct traversal of links directories. Priority: 1
- See Directory monitoring
- Misc
This covers other subtle cases that were made possible by Update Manager.
- Installations are movable when they are self-contained. Priority 1.
- Plug-ins installed through BundleContext.install() are being preserved on restart. Priority 1.
- Startup performance, ensure that simpleconfigurator performs as good if not better than update.configurator. Priority 1.
- Support separation of the eclipse.exe and the configuration folder
- Review the concept of policy files. Priority 1.
Tooling
- Generation of p2 repositories at build time
- The PDE build mechanism produces metadata and artifact repositories as part of the normal build. Priority: 2
- * Application and Ant task to publish content into repositories
- * Application and Ant task to generate metadata / artifacts from existing bundles / features
- * Application and Ant task to optimize a repository
- * Application and Ant task to invoke the mirroring application
- * Modification of the build infrastructure to produce artifact and metadata repos
- [john - optimizer and mirroring support are covered separately below]
- Generation of p2 enabled products at build time
- The PDE build mechanism produces all of the p2 related artifacts and metadata (e.g., install registry, profile, ...) when RCP apps are being built. This allows applications deployed as zips to be p2 enabled out of the box.
- * Modification of the product build to produce p2 eanbled products
- [john - missing priority, and this could also be handled by p2 being able to bootstrap itself on a product in the old shape]
- Streamlined p2 self-hosting
- PDE incrementally and continuously produces p2 metadata and artifact information based on the contents of the workspace and target platform. This simplifies the development of p2-enabled applications by eliminating the need for time-consuming deployment and export cycles while testing and allowing developers to install bundles directly from their workspace without exporting or deploying. Priority: 2
- * PDE Model watcher invoking the metadata generation facility and generate artifacts
- * Provision on launch
- * Target provisioner reading the bundles.txt
- * Definition of a metadata generator advise format, see if this somewhat relates to an externalized format of the IU.
- Provisioning the target
- PDE's Target Provisioner mechanism has been extended to allow the use of p2 when adding bundles to the target platform. This allows bundle developers to benefit from all facilities in p2 when managing their targets. Priority: 2
- * Profile browser
- * Discovery of the profile and agent by which an eclipse is managed
- * Workflow / User UI to add IUs to the target
- Repository browsers and editors
- p2 tooling includes browsers and editors for the artifact and metadata repositories. Users can view, add and remove elements from local and remote p2 repositories. Priority: 1
- * Provide viewers to browse repo content
- * Provide addition / removal capabilities of item into repository
- * Provide IU / artifact correlation tool to check if all the artifacts of an IU are available in an artifact repository, to remove all the artifacts related to an IU, etc.
- [john - I suggest p2]
- Metadata Authoring
- p2 tooling will offer the ability to author installable units. Priority 3.
- * Define a serialized format of IUs for edition purpose.
- * Define an editor and a model allowing for ease of edition.
- * ...
- Migration tools
- Developers can deploy existing features and Eclipse product configurations into p2 repositories using p2 Publisher tools that automatically transform runtime and Update Manager markup into p2 data structures and add this data and artifacts to the relevant repositories. Priority: 1
- [john - we don't need this if p2 can bootstrap itself when dropped in a legacy application in the legacy shape (akin to on the fly generation of MANIFEST.MF for old plugins)]
- Repository Optimizers
- p2 includes tools that analyze, transform and optimize the artifacts in an artifact repository to improve download time, enhance security, etc. Priority: 1
- * Ability to publish artifacts with known processing steps
- * Application and Ant task to optimize a repository
- [john - this item is quite open-ended. To me this is p2 or p3, unless there are some very specific targeted tools that we really that can be marked p1]
- Mirroring tools
- Artifact and metadata repositories can be duplicated in whole or in part using a set of tools included in p2. Priority: 1
- * Application and Ant task to invoke the mirroring application
- [john - I suggest p2 or p3. Isn't rsync all we need?]