1 Abstract

In 2023, Inline XBRL text block tagging became mandatory for European Single Electronic Filing (ESEF) reports. Unlike previous Inline XBRL reporting systems such as UK Companies House and US SEC reporting, ESEF reports often make use of heavily-styled XHTML designed to replicate the layout and appearance typically seen in PDF financial reports.

The ability of Inline XBRL to provide a single document that combines both human-readable content and machine-readable data into a single document, without compromising on the appearance of the report is very powerful and opens the door to the adoption of Inline XBRL in a wider range of reporting systems.

The use of heavily-styled XHTML creates some specific technical issues for text block facts. The extent to which the rendering of the XHTML fragment contained in the value of a text block fact extracted from an Inline XBRL report should replicate the appearance of the same content appearing in the Inline XBRL has been a source of concern for preparers and reviewers alike.

This working group note discusses the technical challenges with Inline XBRL text block tags, and makes a number of recommendations on how to address them.

2 Block tagging

Block tagging (or "text block tagging") refers to the practice of tagging sections of text content in an XBRL or Inline XBRL report. Block tags can be used to tag single sentences, paragraphs, or even full reports, including images, tables and any other content.

XBRL taxonomy concepts used for block tags usually have a data type of textBlockItemType (as defined in the DTR), and the content of a block tag is a fragment of XHTML, meaning that the content can contain formatting instructions.

This document discusses the issues currently seen when applying block tags to ESEF reports. This can be categorised into two broad categories:

1. Display issues – technical issues with rendering fragments of XHTML
2. Content issues – issues related to the scope of block tagging required

These are discussed separately.

3 Display issues

3.1 Block tagging in Inline XBRL

The Inline XBRL specification provides a mechanism (the escape="true" setting) for tagging part of an Inline XBRL report such that value of the extracted XBRL fact is an XHTML fragment containing the contents of the Inline XBRL tag.

Although the Inline XBRL specification defines how these XHTML fragments are obtained from the source document, it does not say anything about how these extracted XHTML fragments are to be used or displayed, and this lack of specification is at the heart of many of the display issues seen with block tags.

There are many potential issues that can occur when extracting a fragment of XHTML from a larger document and attempting to display it. For example:

• The fragment may use CSS classes for styling. These rely on a separate stylesheet that does not form part of the extracted fragment, meaning that the resultant rendering may look quite different to the original.

• Even if the fragment uses inline CSS styling, the styling may rely on the styling of enclosing XHTML elements that do not form part of the fragment.

  As a simple example, the text may use a white font colour because it is displayed on a dark background, but the element with the dark background is not included in the fragment. If the extracted XHTML fragment is then displayed against a white background, the text will be unreadable.

  Another example is an XHTML element that uses CSS positioning to position the element relative to some ancestor. If that ancestor is not included in the fragment, it is unlikely to display correctly.

• The fragment may include images using relative URIs. There is no guarantee that images will be available at the same URIs if the fragment is displayed.

In this document, we will use the term isolated rendering to refer to the display of the fragment of XHTML in a block tag in isolation from the XHTML document from which it was drawn.

3.2 Block tags at the SEC

Prior to the introduction of block tagging as part of the European Single Electronic Filing (ESEF) programme, the most notable use of block tags has been in filings to the US SEC. The use of block tags at the SEC has been successful as the potential issues with displaying extracted fragments of XHTML discussed above are not a problem in practice. There are a number of reasons for this:

1. SEC filing rules mandate the use of inline CSS styles. This removes the issue of CSS class definitions being unavailable when displaying the text block. The use of inline CSS styles is prescribed by the Edgar Filer Manual, and enforced by validation.

2. The styling applied to Inline XBRL documents filed to the SEC is typically quite simple, and does not make use of CSS positioning to achieve complex document layouts, custom fonts, or heavily-styled table backgrounds and borders.

3. The SEC provides viewer software that provides a de-facto specification of how text block content is to be displayed.

   The software is freely available, allowing preparers (and their software providers) to ensure that block tags display correctly in this software.

An extract from a typical SEC Inline XBRL report is shown in Figure 1.

Figure 1: Extract from typical SEC Inline XBRL report

An example of a text block tag shown in the SEC's viewer is shown in Figure 2.

Figure 2: Text block tag shown in SEC viewer

3.3 Block tags in ESEF reports

The ESEF programme (and the closely related UK-equivalent, UKSEF²) has many similarities with the SEC's, but the nature of the XHTML in ESEF Inline XBRL reports is often very different to that of SEC filings. ESEF reports commonly make use of heavily-styled XHTML designed to replicate the layout and appearance typically seen in PDF financial statements.

An example of a page from a typical ESEF report is shown in Figure 3.

Figure 3: Example of a typical, heavily-styled ESEF report

Common features of such reports include:

• A fixed layout that does not respond to the size of the display window.

• A paginated layout, replicating the distinct pages common in a PDF report.

• Fixed text layout. Individual lines within a paragraph are absolutely positioned, and words do not re-flow onto different lines if the window is resized.

• Use of custom fonts, and varying text sizes and colours.

• Heavy use of images, including background images that provide page formatting such as table lines and borders.

ESEF filings are not restricted to using inline styles. It would be impractical to achieve the appearance of these documents with such a restriction.

These features can create issues when attempting to display an XHTML fragment. An extreme example of this is shown below.

Figure 4 shows the first part of a text block tag taken from an ESEF report.

Figure 4: Example text block tag (source rendering)

Figure 5 shows an isolated rendering of the same block tag. As can be seen, the result is unusable. This is due to inline CSS styles that make use of absolute positioning that relies on container elements that are not included in the text block tag; rather than being positioned relative to separate page containers, they are all positioned relative to the same container, resulting in overlapping text.

Figure 5: Example text block tag (isolated rendering)

3.3.1 ESEF report production

Many ESEF reports are currently prepared by converting PDF reports into XHTML, and then add Inline XBRL tags. The PDF conversion process attempts to exactly replicate the appearance of a PDF, and this approach invariably results in the fixed layout features noted above. It also tends to produce very large XHTML documents, as many additional XHTML tags are included in order to precisely position individual lines, words, or even letters.

There is a growing market of XHTML-native report preparation systems that target the creation of XHTML directly. These can often achieve the same high level of design fidelity but with much simpler and cleaner XHTML. Display of block tags drawn from such documents is usually much better than with PDF-converted documents, but is seldom perfect.

It is important to recognise that the use of PDF-conversion preparation solutions has been essential to the success of ESEF. If preparers had not been able to straightforwardly replicate the appearance of their existing PDF reports, it is very unlikely that ESEF's Inline XBRL reports would have been accepted in place of PDF.

3.4 Understanding reporting requirements

It should be noted that the differences described above reflect differences in the underlying reporting requirements. The use of relatively plainly-styled HTML in SEC submissions pre-dates the adoption of XBRL. Similarly, the use of heavily-styled "glossy" annual financial reports has been the norm in many parts of Europe prior to ESEF.

There are also many Inline XBRL reporting systems that do not make of use text block tagging at all.

Inline XBRL is also increasingly used for Environmental, Social and Governance (ESG) reporting. ESG reports typically make heavy use of narrative disclosures that are captured in XBRL as text block tags.

A major strength of Inline XBRL is the ability to support, rather than disrupt, existing human-readable reporting practice and combine it with structured, machine-readable data. In evaluating the applicability of the recommendations of this Working Group Note to a given reporting environment, it is important to fully understand the underlying reporting requirements.

3.5 Use cases for text block tags

Addressing the display issues seen with block tags in ESEF reports requires a new approach. In order to define this approach, we should first consider what block tags are useful for.

It is often assumed that the primary purpose of block tags is isolated rendering — displaying the tag content separately from the source document — but this assumption should be verified.

The use of block tags in SEC filings pre-dates the adoption of Inline XBRL. With an xBRL-XML (XBRL v2.1) report, being able to display the contents of a block tag is critical, as it is the only way to see the content of textual disclosures.

With Inline XBRL, tags are linked to the source XHTML document. Block tags can be used to navigate to, and highlight, the relevant content in the source document, preserving all original styling. Figure 6 demonstrates how the labels in an XBRL taxonomy can be used to search and navigate an Inline XBRL report. In this example, English language labels are used to search a French language report.

Figure 6: Example of using taxonomy labels to navigate an Inline XBRL report

Figure 7 shows how Inline XBRL tags can be used to focus on relevant content in a report.

Figure 7: Example of highlighted text block tag

The ability to view block tags in the context of the source document provided by Inline XBRL greatly reduces the importance of an isolated rendering that exactly replicates the appearance of the source document.

3.5.1 Side-by-side comparison

Isolated rendering of block tags may actually be more useful if the source styling is removed. As an example, consider an analyst wanting to compare facts for the same concept reported by different companies by displaying them side-by-side.

Figure 8 and Figure 9 show facts highlighted in two different ESEF reports. In both cases, the block tag spans multiple pages, and in the first case, the report uses a landscape layout with multiple columns.

Figure 8: Source report 1, with relevant tag content highlighted

Figure 9: Source report 2, with relevant tag content highlighted

Both examples use a fixed layout; the CSS does not allow the content to be re-flowed to a different line length.

In this case, the source styling hinders a side-by-side comparison of the reports as an unreadably small font size is required if the layout is retained (Figure 10).

Figure 10: Side-by-side view using source styling

It is not possible to render this content, with its styles, outside of the context of the enclosing page, as each line of text uses absolute positioning relative to the enclosing page container.

If instead we remove the source styling, the content can be re-flowed and re-styled to achieve a much more useable result:

Figure 11: Side-by-side view using unstyled structured text

Achieving this result requires that the documents contain appropriate semantic XHTML tags (see Section 5.1.1).¹

3.6 Supported use cases

Block tags in Inline XBRL reports should support the following use cases:

1. Disclosure indicator. The presence of a block tag serves as an indicator that a report contains a fact for that concept. This allows consumers to quickly find reports containing a given disclosure. This use case would be satisfied even if the block tags are not connected with the XHTML at all, for example, by including them as hidden tags, but this approach is neither proposed nor recommended.
   

2. Navigation and highlighting. Block tags allow a user to find and see the tagged content for a given fact in the context of the source document. Navigation can make use of the rich metadata provided by an Inline XBRL report and its accompanying taxonomy including multi-language labels, and references to authoritative standards.

3. Extraction of structured text. Block tags should allow a user to reliably extract text that preserves just text structure (headings, paragraphs, lists, and tables), and which can be styled and re-flowed by the consumer. This facilitates usage such as the side-by-side view comparison described in Section 3.5.1.

In some cases, there may be little value in supporting the extraction of structured text. This is discussed in Section 4.2.

3.7 Unsupported use cases

Block tags in Inline XBRL reports should not attempt to support the isolated rendering of an extracted fragment of XHTML in a manner that retains all source styling and layout.

As noted above, the need for an isolated rendering of block tags is largely redundant with Inline XBRL due to the availability of the tagged source document. Use cases that might otherwise require this are adequately (or better) served by navigation and highlighting, or by the extraction of structured text.

As discussed in Section 3.3, there are significant technical challenges with achieving an isolated rendering of a fragment of XHTML taken from a heavily-styled XHTML document. Indeed, there are significant challenges in even specifying what this would mean. For example, where a block tag is split across multiple pages and columns, is it necessary to maintain the relative positioning of the components of the tag in the isolated rendering?

4 Content issues

ESEF filing requirements prescribe a large number of block tag concepts that must be tagged if they are present in the notes to the financial statements in a report. Many of these mandatory concepts overlap, with more specific concepts included that form part of larger, less-specific concepts. The net result of this is that some parts of the report may be tagged many times using nested block tags.

Figure 12 shows an example of a section of a report that has been included as part of 13 different block tags.

Figure 12: Example of nested multi-taggging

The content of some block tag concepts is very large, often including many pages from the report.

The combination of nested tagging, tags that include large parts of the source report, and the inefficient XHTML seen in PDF-converted reports (see Section 3.3.1) can lead to extremely large XBRL reports when extracting data from the Inline XBRL.

4.1 Nested block tags and multi-tagging

There are different views on the benefits of heavily nested tagging. On the one hand, tagging content with all applicable concepts makes it easy for consumers to directly access the content for a given concept. On the other hand, nested tagging increases preparer burden and increases extracted report size, and in many cases consumers can infer a larger enclosing tag from its individually tagged components.

4.2 Large block tag use cases

As noted above, some block tags cover large sections of the report. The Working Group considers that these large section-level tags are only useful for the disclosure indicator and navigation and highlighting use cases described in Section 3.6; the utility of extracting structured text from such section-level tags is limited or non-existent, given that all of the content will typically also be tagged by more granular block tags.

5 Proposed solution

This note proposes two approaches for addressing the issues described above:

• Structured text extraction
• No-content block tags

These solutions are outlined below.

5.1 Structured text extraction

As described in Section 3.6, when extracting content from a block tag, the goal should be to preserve text structure, but not styling. In order to facilitate this, it is proposed that new Inline XBRL transformations are introduced that remove styling information and purely presentational XHTML elements (<div> and <span>), and leave only unstyled, structural XHTML elements (see Section 5.1.1).

Consumers already have the ability to strip non-structural XHTML elements and styling today, but as this process is not defined or standardised, report preparers cannot reasonably target this output when reviewing the output of block tags. Many reports do not have the necessary structural XHTML elements for this process to be effective.

By introducing a standardised transformation, preparers will have a clear specification of how block tags will be handled, and can therefore focus on ensuring that block tags produce correctly structured structural XHTML, and avoid expending effort on the effect of styling instructions once they are removed from the context of the surrounding document.

Further, specifying this HTML simplification process as a transformation should significantly reduce the size of the extracted fact values, as the presentational <div> and <span> elements are removed as part of the extraction.

5.1.1 Structural XHTML tags

The XHTML tags that control the appearance of an Inline XBRL report can be created in different ways. It is possible to use structural XHTML tags, such as <p> (paragraph), <h1> (level 1 heading) and <table> (table) that provide information about the structure of the document.

Alternatively, presentational tags (<div> and <span>) may be used. The application of CSS styling can be used to recreate exactly the same visual appearance under either approach, but the information available to consumers about the structure of the document is different.

The XHTML specification refers to structural tags as "semantic tags" (see section 3.2.1 of the XHTML standard), but this document uses the term "structural" in order to avoid confusion with the semantics provided by Inline XBRL tags.

Although the Inline XBRL specification does not require the use of structural XHTML tags, there are good reasons to prefer semantic XHTML tags, where possible. For example, semantic XHTML tags improve the accessibility of a report, and can make it easier to navigate. XHTML table tags can be used to capture the row/column structure of a table, making it easier to copy tabular data to other formats (e.g. Excel spreadsheets).

5.1.2 XHTML semantics vs Inline XBRL semantics

XHTML tagging, even with structural XHTML tags, is not a substitute for Inline XBRL tagging of individual facts. XHTML semantics relate to the structure of the document; Inline XBRL tagging provides additional semantics about the facts being reported such as units, period, precision, dimensions and scaling, and most importantly, connections to a taxonomy that provides the semantic definition of the concepts and dimensions.

This is particularly important for tables, where detailed tagging of numerical data is considered to be the most useful and will provide detailed information about each number being disclosed, whereas XHTML semantics only convey the row and column structure.

5.1.3 Combining structural XHTML with presentational XHTML

Many ESEF documents do not currently contain the necessary structural XHTML elements required under this approach. These would need to be added as part of the document preparation process. Some preparation software already provides the ability to do this with tables (inserting <table>, <tr> and <td> tags into tables that are laid out using absolutely positioned <div> and <span> elements). It should be possible for structural XHTML elements to co-exist with the presentational XHTML that the documents currently rely on. For example, reports may be able to use the display: contents CSS instruction on structural XHTML elements to remove any impact on the source document's appearance.

5.1.4 Improving structural XHTML output

The proposed transform could also provide a mechanism for improving the XHTML structural information. For example, where a report uses a fixed layout with <div> elements for pages and columns, a paragraph that spans a column or page break cannot be marked up with a single <p> tag.

For example, the following is invalid HTML, because the <p> tag is not nested correctly within a single parent:

    <div class="page">
        <p>
            This paragraph starts on the first page...
    </div>
    <div class="page">
        ... and continues on the next page.
        </p>
    </div>

This would need to be structured as follows, using separate <p> tags:

    <div class="page">
        <p>
            This paragraph starts on the first page...
        </p>
    </div>
    <div class="page">
        <p>
        ... and continues on the next page.
        </p>
    </div>

This leads to an unwanted paragraph break between the two parts of the paragraph. The transform could provide a mechanism for constructing a single <p> tag as part of the block tag extraction process. The details of this mechanism would need to be defined, but it could, for example, use CSS classes to indicate that a paragraph follows on from the previous paragraph.

5.2 No-content block tags

As described in Section 4.2, large, section-level block tags are most valuable for the disclosure indicator and navigation and highlighting use cases. Therefore, it is proposed that an approach for tagging such block tags in a way that produces no extracted content is standardised.

This approach would retain support for the two use cases mentioned above, but would avoid bloating the extracted XBRL report with very large block tag content.

There are existing transformation rules that could be used for this purpose. For example, ixt:fixed-empty and ixt:fixed-true transform any input content to an empty string, or the string true, respectively, but it may make sense to introduce a transform specifically for this purpose.

5.3 Tagging approach selection

The choice of which approach should be used for a given tag would need to be prescribed by the relevant filing authority.

The implementation of these proposals (see Section 6) should provide an explicit mechanism, such as standardised datatypes, for indicating which approach is to be used for each concept.

5.4 Nested tags

Both of the approaches described above would reduce the extracted report size, and should reduce preparer burden by simplifying, or removing, the content of extracted block tags. As such, these go some way to addressing the issues with heavily nested multi-tagging with block tags. This working group note does not take a position on whether the number of nested block tags required in ESEF reports should also be reduced.

6 Next steps

This document aims to describe and explain the issues currently seen with block tags in some Inline XBRL reports, and provides a high-level description of potential solutions.

Adoption of these proposed solutions would require a more formal definition in the form of additional transformation rules and additional datatypes.

The definition of exactly which XHTML tags are included in the extracted output will be prescribed as part of the definition of the transformation rules.

The formal definition of transformation rules and additional datatypes will be accompanied by best practice guidance documenting their use.

Once these components have been defined, updates to filing rules specifying their use will be required.

Appendix A Semantic XHTML and PDF-based preparation

Many ESEF reports are currently prepared using a process that converts a PDF document to XHTML. Such a conversion process will not automatically produce semantic XHTML tags, but will instead produce positioned <div> and <span> elements that recreate the appearance of the original PDF document. In such systems, additional effort may be required in order to insert structural XHTML tags. A number of software vendors have indicated willingness to implement features that would minimise additional effort required by preparers.

A requirement to produce documents that use structural XHTML tags exclusively may impose an unacceptable disruption to existing preparation processes. The approach described in Section 5.1 would allow structural XHTML tags to co-exist with such presentational XHTML, providing the benefits of structural XHTML tags without removing the ability to use presentational XHTML tags to recreate a fixed layout.