Microsummaries

Introduction

Microsummaries are regularly-updated succinct compilations of the most important information on web pages. They are compact enough to fit in the space available to a bookmark label, provide more useful information about pages than static page titles, and are regularly updated as new information becomes available.

Microsummaries are better than page titles at labeling bookmarks because they give users quicker access to the most interesting information behind a bookmark and provide sites with a way to notify users of updates and entice them to revisit the site.

For a quick overview of microsummaries and their proposed integration into Firefox, see this walkthrough with screenshots showing major elements. For detailed information, read the rest of this document.

Examples

Examples of pages for which microsummaries are useful include:

• auction items: the item name, the current highest bid, and the time remaining, f.e. Honda Accord - $5000 - 1 minute left;
• products for sale: the product name, the current price, and whether or not the product is in stock, f.e. Linksys WRT54G - $60 - in stock;
• news sites: the latest headline, f.e. BBC: Chirac to sign France's job law;
• "[thing] of the day" pages: today's [thing], f.e. (for Merriam-Webster's word of the day page) flat-hat;
• stock quotes: the current price of the stock and its movement in the market, f.e. TWX 16.94 + .30;
• stock portfolios: your current net worth, f.e. net worth: $30k; +$500 today;
• weather pages: the current forecast, f.e. SF: showers likely;
• tinderbox: the status of the tree, f.e. 3 tbxn burning;
• insert your bright idea here.

Microsummary Generators

Microsummaries can either be provided by the page being summarized or generated by processing an XSLT stylesheet against the page. In the latter case, the XSLT stylesheet and the page(s) to which it applies are provided by a microsummary generator.

A microsummary generator is a set of instructions for generating microsummaries. Generators are expressed via an XML dialect with the namespace http://www.mozilla.org/microsummaries/0.1. A generator consists of a <generator> tag containing the following attributes and child elements:

• a name attribute that identifies the generator to users, f.e. Latest BBC Headline;
• a <pages> child element with one or more <include>/<exclude> child elements containing regular expressions that identify the URIs to which the generator applies; URIs are excluded by default and must be explicitly included for a generator to apply to them;
• a <template> element containing an XSLT stylesheet which generates a microsummary from a page. The stylesheet can generate not only text but also HTML, graphics, and other rich content.

[Should a generator also contain an update interval or timeout specifying how often to update microsummaries or when to update them next? How should this interval/timeout be specified? Is it worth enabling complex specifications like "update every five minutes from 9:30am-4pm EST and not at all otherwise" for an NYSE stock quote?]

Here is an example microsummary generator:

 <?xml version="1.0" encoding="UTF-8"?>
 <generator xmlns="http://www.mozilla.org/microsummaries/0.1" name="Yahoo! Finance Stock Quote">
   <pages>
     <include>http://finance.yahoo.com/q\?.*s=[a-zA-Z]+</include>
   </pages>
   <template>
     <xsl:transform xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0">
       <xsl:output method="text"/>
       <xsl:template match="/">
         <xsl:value-of select="substring-before(substring-after(string(id('yfncsubtit')/tbody/tr/td[1]/b), '('), ')')"/>
         <xsl:text>: </xsl:text>
         <xsl:value-of select="id('yfncsumtab')/tbody/tr/td/table[3]/tbody/tr[2]/td[1]/table/tbody/tr/td/table/tbody/tr[1]/td[2]/big/b"/>
       </xsl:template>
     </xsl:transform>
   </template>
 </generator>

Specifying Microsummaries

Firefox, its users, the web sites they visit, and independent developers should all be able to specify microsummaries for pages.

Firefox

Firefox should be able to specify microsummaries via bundled microsummary generators for popular sites.

Users

Users should be able to specify microsummaries via a microsummary builder feature in Firefox that lets users extract text snippets from a rendered page via simple methods like drag-and-drop and string them together along with arbitrary text into a microsummary. Implementation of a microsummary builder is currently deferred to the future.

Sites

Sites should be able to specify microsummaries by embedding metadata referencing microsummary generators into pages. For example, a site might embed the following <link> element into an HTML page:

 <link rel="microsummary" type="application/x.microsummary+xml" href="/index-microsummary.xml">

Or it might embed the following two XML processing instruction into an XML document:

 <?microsummary type="application/x.microsummary+xml" href="/index-microsummary.xml"?>

[If a page points to a microsummary generator whose <pages> list does not include the page, should the microsummary service ignore the <pages> list and apply the generator to the page anyway?]

Independent Developers

Independent developers should be able to specify microsummaries by distributing generators that users can acquire from a directory similar to the search engines directory at addons.mozilla.org. Generators should have simple search engine-like installation and management UI and update themselves automatically.

Back-end Implementation

The microsummary service updates microsummaries when they expire and provides an API for front-end code to access microsummaries and be notified when they get updated.

Like the livemarks service, the microsummary service checks every 15 seconds for microsummaries that need updating. If a microsummary needs updating, the service downloads the necessary content (i.e. the page, the generator), processes it to generate an updated microsummary, and stores the updated microsummary in the datastore.

nsIMicrosummaryService

 [scriptable, uuid(c5e9c390-beb0-4eb4-90ab-529efc817632)]
 interface nsIMicrosummaryService : nsISupports
 {
     /**
      * Install the microsummary generator from the resource at the supplied URI.
      * Callable by content via the addMicrosummaryGenerator() sidebar method.
      *
      * @param   generatorURI
      *          the URI of the resource providing the generator
      *
      */
     void addGenerator(in nsIURI generatorURI);
 
     /**
      * Get the set of microsummaries available for a given page.  The set
      * might change after this method returns, since this method will trigger
      * an asynchronous load of the page in question (if it isn't already loaded)
      * to see if it references any page-specific microsummaries.
      *
      * If the caller passes a bookmark ID, and one of the microsummaries
      * is the current one for the bookmark, this method will retrieve content
      * from the datastore for that microsummary, which is useful when callers
      * want to display a list of microsummaries for a page that isn't loaded,
      * and they want to display the actual content of the selected microsummary
      * immediately (rather than after the content is asynchronously loaded).
      *
      * @param   pageURI
      *          the URI of the page for which to retrieve available microsummaries
      *
      * @param   bookmarkID (optional)
      *          the ID of the bookmark for which this method is being called
      *
      * @returns an nsIMicrosummarySet of nsIMicrosummaries for the given page
      *
      */
     nsIMicrosummarySet getMicrosummaries(in nsIURI pageURI,
                                          in nsISupports bookmarkID);
 
     /**
      * Get the current microsummary for the given bookmark.
      *
      * @param   bookmarkID
      *          the bookmark for which to get the current microsummary
      *
      * @returns the current microsummary for the bookmark, or null
      *          if the bookmark does not have a current microsummary
      *
      */
     nsIMicrosummary getMicrosummary(in nsISupports bookmarkID);
 
     /**
      * Set the current microsummary for the given bookmark.
      *
      * @param   bookmarkID
      *          the bookmark for which to set the current microsummary
      *
      * @param   microsummary
      *          the microsummary to set as the current one
      *
      */
     void setMicrosummary(in nsISupports bookmarkID,
                          in nsIMicrosummary microsummary);
 
     /**
      * Remove the current microsummary for the given bookmark.
      *
      * @param   bookmarkID
      *          the bookmark for which to remove the current microsummary
      *
      */
     void removeMicrosummary(in nsISupports bookmarkID);
 
     /**
      * Whether or not the given bookmark has a current microsummary.
      *
      * @param   bookmarkID
      *          the bookmark for which to set the current microsummary
      *
      * @returns a boolean representing whether or not the given bookmark
      *          has a current microsummary
      *
      */
     boolean hasMicrosummary(in nsISupports bookmarkID);
 
     /**
      * Whether or not the given microsummary is the current microsummary
      * for the given bookmark.
      *
      * @param   bookmarkID
      *          the bookmark to check
      *
      * @param   microsummary
      *          the microsummary to check
      *
      * @returns whether or not the microsummary is the current one
      *          for the bookmark
      *
      */
     boolean isMicrosummary(in nsISupports bookmarkID,
                            in nsIMicrosummary microsummary);
 };

nsIMicrosummaryObserver

 [scriptable, uuid(cb284a83-1ca5-4000-9841-ce345ce84915)]
 interface nsIMicrosummaryObserver : nsISupports
 {
     /**
      * Called when an observed microsummary updates its content.
      * Since an observer might watch multiple microsummaries at the same time,
      * the microsummary whose content has been updated gets passed
      * to this handler.
      * XXX Should this be onContentUpdated?
      *
      * @param microsummary
      *        the microsummary whose content has just been updated
      *
      */
     void onContentLoaded(in nsIMicrosummary microsummary);
 
     /**
      * Called when an element is appended to a microsummary set.
      * XXX Should this be in a separate nsICollectionObserver interface?
      *
      * @param microsummary
      *        the microsummary that has just been appended to the set
      *
      */
     void onElementAppended(in nsIMicrosummary microsummary);
 };

nsIMicrosummary

 [scriptable, uuid(f9d1a73c-e147-46f3-ba61-4f0bd33f5d47)]
 interface nsIMicrosummary : nsISupports
 {
     // the URI of the page being summarized
     readonly attribute nsIURI pageURI;
 
     // The URI of the generator that generates this microsummary.
     readonly attribute nsIURI generatorURI;
 
     // The generator that generates this microsummary.
     // Since generators can be remote resources, this may not always be available.
     attribute nsIMicrosummaryGenerator generator;
 
     // The content of the microsummary.
     // Since generators and pages can be remote resources, and we need them
     // to generate the content, this may not always be available.
     readonly attribute AString content;
 
     /**
      * Add a microsummary observer to this microsummary.
      *
      * @param observer
      *        the microsummary observer to add
      *
      */
     void addObserver(in nsIMicrosummaryObserver observer);
 
     /**
      * Remove a microsummary observer from this microsummary.
      *
      * @param observer
      *        the microsummary observer to remove
      *
      */
     void removeObserver(in nsIMicrosummaryObserver observer);
 
     /**
      * Update the microsummary, first loading its generator and page content
      * as necessary.  If you want know when a microsummary finishes updating,
      * add an observer before calling this method.
      *
      */
     void update();
 };

nsIMicrosummarySet

 [scriptable, uuid(7111e88d-fecd-4b17-b7a9-1fa74e23153f)]
 interface nsIMicrosummarySet : nsISupports
 {
     /**
      * Add a microsummary observer to this microsummary set.  Adding an observer
      * to a set is the equivalent of adding it to each constituent microsummary.
      *
      * @param observer
      *        the microsummary observer to add
      *
      */
     void addObserver(in nsIMicrosummaryObserver observer);
 
     /**
      * Remove a microsummary observer from this microsummary.
      *
      * @param observer
      *        the microsummary observer to remove
      *
      */
     void removeObserver(in nsIMicrosummaryObserver observer);
 
     /**
      * Retrieve a enumerator of microsummaries in the set.
      *
      * @returns an enumerator of nsIMicrosummary objects
      *
      */
     nsISimpleEnumerator Enumerate();
 };

nsIMicrosummaryGenerator

 [scriptable, uuid(ff3eba15-81de-4c24-bfcf-c8180dc3c00a)]
 interface nsIMicrosummaryGenerator : nsISupports
 {
     // An arbitrary descriptive name for this microsummary generator.
     readonly attribute AUTF8String name;
 
     // The canonical location and unique identifier of the generator.
     // It tells us where the generator comes from and where to go for updates.
     // For generators referenced by web pages via <link> tags, this URI
     // is the one specified by the tag.  For generators installed by users,
     // this URI is the remote location from which we installed the generator.
     // For generators bundled with the browser, this URI is a reference
     // to the latest version of this generator available on Mozilla Add-ons.
     readonly attribute nsIURI uri;
 
     // Whether or not this generator needs page content to generate
     // a microsummary. Microsummaries generated by XSLT templates need page
     // content, while those which represent the actual microsummary content
     // do not.
     readonly attribute boolean needsPageContent;
 
     /**
      * Generate a microsummary by processing the generator template
      * against the page content.  If a generator doesn't need content,
      * pass null as the parameter to this method.
      *
      * XXX In the future, this should support returning rich text content,
      * so perhaps we should make it return a DOM node now and have the caller
      * convert that to text if it doesn't support rich content.
      *
      * @param   pageContent
      *          the content of the page being summarized
      * @returns the text result of processing the template
      *
      */
     AString generateMicrosummary(in nsIDOMNode aPageContent);
 };

Datastore

In Firefox 2 the microsummary service stores microsummaries and their meta-data as properties of the RDF resources that represent bookmarks in the bookmarks data source. In Firefox 3 the service stores the same date as annotations in the annotations datastore via the annotation service. Microsummary data includes:

• bookmarks/generatedTitle: the content of the microsummary (i.e. what Firefox displays to users);
• microsummary/generatorURI: the canonical source of the generator;
• microsummary/expiration: the time in microseconds since the epoch at which the microsummary will expire.

Front-end Implementation

The initial integration points are the bookmark properties dialog and the bookmarks toolbar.

Since microsummaries can contain HTML and other unsafe content, they should be inserted into the UI inside untrusted iframes wherever they appear.

Bookmark Properties Dialog

The bookmark properties dialog lets the user choose to display a microsummary for the bookmark. If multiple microsummaries are available, the dialog lets users choose between them. The dialog retrieves microsummaries via nsIMicrosummaryService.getMicrosummaries() and updates the datastore per the user's selection via nsIMicrosummaryService.setMicrosummary() and nsIMicrosummaryService.removeMicrosummary().

Bookmarks Toolbar

When a microsummary is set for a bookmarked URI, the bookmarks toolbar displays the microsummary instead of the page title as the label of the bookmark. When the microsummary service updates the microsummary, the bookmarks toolbar updates the label. Once the microsummary service has updated the microsummary, the bookmarks toolbar may display some UI indicating that an update is underway (f.e. it may replace the favicon with a throbber for the duration of the update).

The toolbar displays microsummaries via a template rule that applies only to bookmarks with microsummaries. The toolbar controller registers itself with the microsummary service as a microsummary observer in order to be notified when a microsummary gets updated.

[The mechanism by which the bookmarks controller identifies and observes microsummary bookmarks should be extensible (i.e. a generic "metadata observer") so that future code (both native and extensions) can register additional bookmark types with metadata whose observation triggers activity.]

Additional integration points may be defined in the future (f.e. microsummaries might be displayed as tab labels).

Standardization

The microsummary generator dialect and the use of the <link rel> element to specify microsummaries should be standardized by appropriate bodies. Appropriate bodies for standardization may include the microformats group and the WHATWG.