HTML Imports are a way to include and reuse HTML documents in other HTML documents [[!HTML]].

Any point, at which a conforming UA must make decisions about the state or reaction to the state of the conceptual model, is captured as algorithm. The algorithms are defined in terms of processing equivalence. The processing equivalence is a constraint imposed on the algorithm implementers, requiring the output of the both UA-implemented and the specified algorithm to be exactly the same for all inputs.

Terminology

HTML Imports, or just imports from here on, are HTML documents [[!HTML]] that are linked as external resources from another HTML document. The document that links to an import is called an import referrer. For any given import, an import referrer ancestor is its import referrer or any import referrer ancestor of its import referrer. There are one or more import referrers and import referrer ancestors for each import because same import can be referred from multiple import referrers.

An import referrer that is not an import, thus is not associated with any import referrer, is called a master document. Each import is associated with one master document: if the referrer of the import is a master document, it is the master document of the import. Otherwise, the master document of the import referrer is the master document of the import.

The URL of an import is called the import location.

In each import referrer, an import is represented as a Document [[!WHATWG-DOM]], called the imported document.

The imported documents don't have browsing context.

The set of all import associated with the master document forms an import map of the master document. The maps stores import as its items with their import locations as keys. The import map is empty at beginning. New items are added to the map as import fetching algorithm specifies.

Import Dependent

To track requested imports, each document has an import link list. Each of its item consists of link, a link element and location, a URL. Also, the item is optionally marked as branch. The list is initially empty, and items are added to it as specified by the import request algorithm.

Each imported document has an import parent: If the import link list of document A contains a branch item whose location points document B, A is an import parent of B.

Each imported document also has one or more import ancestors: Document A is an import ancestor of another document B if A is import parent of B. Being an import ancestor is transitive: If A is an import parent of B and B is an import parent of C, A is an import parent of C as well.

An imported document also has one or more import predecessors. An import predecessor is a document. If the URL of document A is located before the URL of document B in the import link list of B's import parent, and the located link is marked as a branch, then A is import predecessor of B.

The import ancestor predecessors of document A is defined as follows: If document B is an import predecessor of document C, and C is an import ancestor of A, B is an import ancestor predecessors of A.

The Document that is in either import ancestor predecessors or import predecessors of document A, or is linked from branch item of A's import link list, is the import dependent of A.

The import link list and the import dependent constrains the order of script execution in imports. It is intend to give a deterministic order of script execution which is defined by the order of link element in each import. The edges of each node is ordered in terms of import link list. The import predecessors selection is aware of the order.

The linking structure of import link lists forms a directed graph. Each node of the graph is a document and its edge is a link. Branches are intended to form a spanning tree of the graph. This tree gives the deterministic order of the script execution.

An example of import link lists

In ,

The difference between the import referrer and the import parent is that import referrer reflects the state of the node tree and that the import parent is built by the algorithm described in this document.

Extensions to HTMLLinkElement Interface

partial interface HTMLLinkElement {
    readonly attribute Document? import;
};

On getting, the import attribute MUST return null, if:

Otherwise, the attribute MUST return the imported document for the import, represented by the link element.

The same object MUST be returned each time.

An import in the context of the Document of an HTML parser or XML Parser is said to be an import that is blocking scripts if the element was created by that Document's parser, or and the element is a link of type import when the element was created by the parser, and the link is not marked as async, and the the import is yet to be completely loaded, and, the last time the event loop has reached step 1, the element was in that Document, and the user agent hasn't given up on that import yet. A user agent MAY give up on an import at any time.

Giving up an import before it loads, even if the import eventually does still load, means that the script might end up operating with incorrect information. For example, if an import registers a custom element and a script relies on the availability of this element, the script will find that this element is unavailable if the user agent gives up early. Implementers have to balance the likelihood of a script using incorrect information with the performance impact of doing nothing while waiting for a slow network request to finish.

A Document has an import that is blocking scripts if there is an import that is blocking scripts in the Document's import dependent. A Document has no import that is blocking scripts if it does not have an import that is blocking scripts as defined in the previous paragraph.

The state of "has an import that is blocking scripts" can change each time an existing import is completely loaded or new import loading is started. HTML parser has changes to unblock it for each of such timings.

Extensions to Document Interface

Additions to document.open() method

Add following step as the first step of the definition:

  1. Throws an InvalidStateError exception if the Document is an import.

Additions to document.write() method

Add following step as the first step of the definition:

  1. Throws an InvalidStateError exception if the Document is an import.

Additions to document.close() method

Add following step as the first step of the definition:

  1. Throws an InvalidStateError exception if the Document is an import.

Loading Imports

Updating Branch

After a link is added to the import link list, the update marking algorithm MUST be run with the master document. which is equivalent to running these steps:

Input
DOCUMENT, the Document
  1. If the DOCUMENT is the master document, unmark branch of all the links in the import link list of DOCUMENT and every import that is associated to DOCUMENT.
  2. Let LIST be an import link list of DOCUMENT.
  3. For each ITEM in the LIST:
    1. Let LOCATION be a location of ITEM.
    2. Let IMPORT be an import whose URL is same as LOCATION.
    3. If there is no other link whose location is same as LOCATION and which is marked as a branch, mark ITEM as a branch.
    4. If ITEM is marked as a branch and IMPORT is not null, invoke update marking algorithm with IMPORT.

Requesting Import

When user agents attempt to obtain a linked import, they MUST also run the import request algorithm, which is equivalent to running these steps:

Input
LINK, the link element that creates an external resource link to the import.
LOCATION, the URL of the linked resource.
  1. If the async attribute of LINK is true, mark LINK as async.
  2. Let DOCUMENT be a document of LINK.
  3. Let LIST be an import link list of DOCUMENT.
  4. Let ITEM be LINK and LOCATION:
    1. Add ITEM at the end of LIST.
    2. Invoke update marking algorithm with the master document.

Fetching Import

All import linked from documents that is the master document or the one in the import map MUST be fetched using the import fetching algorithm described below, instead of the one that HTML specifies to obtain a linked resouce.

The import fetching algorithm MUST be equivalent to running these steps:

Input
LINK, a link element which makes the external resource link to the import.
LOCATION, the import location
Output
IMPORT, the imported document.
  1. If LOCATION is already in the import map:
    1. Let IMPORT be the imported document for LOCATION and stop.
  2. Fetch a resource [[!Fetch]] from LOCATION with request's origin set to the origin of the master document, the mode to CORS and the credentials mode to same-origin.
    1. If fetched response type is error or the response has a header whose name is Content-Disposition:
      • Add LOCATION and null to the import map and stop.
    2. Let IMPORT be a new Document, the document's address of which is LOCATION
    3. Let PARSER be a new HTML parser, associated with IMPORT
    4. Add LOCATION and IMPORT to the import map.
    5. For each task that the networking task source places on the task queue while fetching:
      1. Fill PARSER's input byte stream with the fetched bytes
      2. Let PARSER process the input byte stream with utf-8 [[!WHATWG-ENCODING]] as a known definite encoding
    6. When no more bytes are available:
      1. Queue a task from the networking task source for PARSER to process implied EOF character

All of loaded imports and imports under loading are in the import link list, thus every import which is linked from imports in the list will also be loaded using the import fetching algorithm, with LOCATION be the import location of the import.

The loading attempt MUST be considered successful if IMPORT is not null on the algorithm completion, and failed otherwise.

Every import that is not marked as async delays the load event in the Document.

The link element fires a simple event called load for successful loading attempt. For failed attempt, it fires a simple event named error.

As an import delays the load event, the Document isn't completely loaded until loading attempts of all of its linked imports are finished.

Imports and Content Security Policy

Content Security Policy [[!CSP3]] MUST restrict import loading through the script-src directive.

Each import MUST be restricted by the Content Security Policy of the master document. For example, if Content Security Header Field is sent to an import, the user agent MUST enforce the policy of the master document to the imported document.

Parsing Imports

Parsing behaviour of import is defined as a set of changes to the HTML Parsing.

Additions to Prepare A Script Algorithm

In step 15 of prepare a script algorithm, modify the last part of condition which begins with If element does not have a src attribute to read:

... and the Document of the HTML parser or XML parser that created the script element has a style sheet that is blocking scripts or has an import that is blocking scripts

Additions to Tree Construction Algorithm

At the DOCTYPE part of section 12.2.5.4.1 The "initial" insertion mode, modify text if the document is not an iframe srcdoc document... as follows

if the document is not an iframe src document nor an import...

In sub-condition named Otherwise of condition An end tag whose name is "script" in "text" insertion mode, modify step 3 to read:

  1. If the parser's Document has a style sheet that is blocking scripts or has an import that is blocking scripts or the script's "ready to be parser-executed" flag is not set: spin the event loop until the parser's Document has no style sheet that is blocking scripts and has no import that is blocking scripts and the script's "ready to be parser-executed" flag is set.

Additions to Parsing XHTML Documents

Modify step 3 of steps that run following preparing the script element to read:

  1. Spin the event loop until the parser's Document has no style sheet that is blocking scripts and has no import that is blocking scripts and the pending parsing-blocking script's "ready to be parser-executed" flag is set.

Scripting in Imports

Additions to Script Enabling Criteria

Add following condition to the list of Enabling and disabling scripting criteria:

Additions to document.currentScript

Modify the definition of document.currentScript as follows:

The currentScript attribute, on getting, MUST return the value to which it was most recently initialized in the document or the import map of the document. When the Document is created, the currentScript MUST be initialized to null. If the Document is an imported document, its currentScript is always null.

Style processing with Imports

The contents of the style elements and the external resources of the link elements in import MUST be considered as input sources of the style processing model [[!CSS2]] of the master document.

Order of Appearances and Imports

The order of appearances of declarations [[!CSS-CASCADE-3]] which come from different documents are determined by the import link tree. If node documents of two declarations differ, compare the tree order of these documents in the import link tree. The last one wins.

Events in Imports

Events in import is defined as a set of changes to the HTML Events.

Additions to Event Handlers

Modify the event handler content attribute's script creation criteria by expanding the first paragraph:

When an event handler content attribute is set, if the element is owned by a Document that is in a browsing context or in an import map, …

Custom Element Processing

This specification redefines custom element order to be the sum of [[!CUSTOM-ELEMENTS]]'s custom element order and import tree order, in which the import tree order is scaled so that its lowest value is always larger than the highest possible value of [[!CUSTOM-ELEMENTS]]'s custom element order.

The import tree order of a given custom element of an import link tree is determined by tree order in an import link tree that was flattened by replacing every import link with the content of its imported document.

The highest stable order is the value that is immediately preceding the custom element order of an element in the first encountered import, in tree order, that has not yet completely loaded. If there is no such element, the highest stable order is the highest custom element order in the flattened import link tree. When processing the base element queue, user agents should only invoke callbacks up to the highest stable order, inclusively.

Because imports load asynchronously, we need to divide a sorted element queue into the part where things have settled down (all imports have loaded), and the part where the loading is still happening, and thus the actual sorting order is not yet determined. For example, suppose you have the following document structure:

index.html:

<link rel="import" href="import.html">
...
<me-second></me-second>
...

import.html:

<me-first></me-first>

The order of custom elements in the flattened import link tree is me-first (1), me-second (2). However, it's very likely that the parser will find out about me-second sooner than me-first, since the latter requires loading the import.html. While the network stack is doing its job, the highest stable order stays at the beginning position. When import.html is ready, the order jumps all the way to me-second (2).

Acknowledgments

David Hyatt developed XBL 1.0, and Ian Hickson co-wrote XBL 2.0. These documents provided tremendous insight into the problem of behavior attachment and greatly influenced this specification.

Alex Russell and his considerable forethought triggered a new wave of enthusiasm around the subject of behavior attachment and how it can be applied practically on the Web.

Dominic Cooney and Roland Steiner worked tirelessly to scope the problem within the confines of the Web platform and provided a solid foundation for this document.

The editor would also like to thank Alex Komoroske, Angelina Fabbro, Anne van Kesteren, Boris Zbarsky, Brian Kardell, Daniel Buchner, Edward O'Connor, Eric Bidelman, Erik Arvidsson, Elliott Sprehn, Gabor Krizsanits, Hayato Ito, James Simonsen, Jonas Sicking, Ken Shirriff, Neel Goyal, Olli Pettay, Rafael Weinstein, Scott Miles, Steve Orvell, Tab Atkins, William Chan, and William Chen for their comments and contributions to this specification.

This list is too short. There's a lot of work left to do. Please contribute by reviewing and filing bugs—and don't forget to ask the editor to add your name into this section.