Given this is the old CORBA style, wouldn't it be more consistent to use CORBA::TinyShort and CORBA::UTinyShort, that matches the other types as defined in D.2.1.5

General consensus is to maintain the current mapping

The general consensus is to maintain CORBA::Int8 and CORBA::UInt8 for mapping the int8 and uint8 types, instead of renaming them to CORBA::TinyShort and CORBA::UnsignedTinyShort.

The bitmask mapping allows misuse and is not really a modern API. At least for the users who map a struct to a class the bitmask mapping could for example use a strong enum (mappping now two bitmasks with the same member name will result in non compiling C++ code), but provide the operators ~, |, &, ^, |=, &=, and ^= to make it safe.

Mapping Bitmasks to structs containing an enum and several operators

As suggested in the issue, mapping bitmasks to structs containing the former enum and several operators improves the usability of the mapped type.

The mapping for alternative struct mapping is lacking the constructors, member initialization, non-const accessors, description of how it works when using IDL4 struct inheritance, etc, see the IDL to C++11 language mapping for a full description

Add additional constructors and accessors to alternative mapping for structs

Clause 8.1.1 provides an alternative mapping for structs using the @cpp_annotation and setting struct_mapping to CLASS_WITH_PUBLIC_ACCESSORS_AND_MODIFIERS. In this resolution, we provide extra constructors and accessor methods for usability purposes.

The IDL 4 spec states the default bit_bound is 32 bits the C++ mapping document states:
IDL bitmask declarations shall be mapped to two C++ type names:

• An unscoped enum named <Bitmask>Bits with an explicitly defined underlying type. That underlying type
  is the smallest mapped unsigned integer type that has sufficient bits for the bit_bound of the bitmask:
  uint8_t, for values between 1 and 8; uint16_t for bit_bound values between 9 and 16; uint32_t, for
  values between 17 and 32; and uint64_t for values between 33 and 64. The <Bitmask>Bits enumerators
  are the values defined in the scope of the IDL bitmask, with each enumerator explicitly initialized to its
  corresponding integer value.

IDL4 specifies the default size for Bitmasks

Clause 7.4.13.4.3.3 of IDL 4.2 states that "By default, the size of a bit mask is 32." Therefore, if @bit_bound is unspecified, we should assume the underlying type is uint32_t. It also states that "Two annotations can be used to amend a bit mask definition:" @bit_bound and @position. While an explicit @bit_bound would change the size of the bit mask, as stated in IDL 4.2, "Possible positions range from 0, which corresponds to the less significant bit, up to (size – 1), which corresponds to the most significant one." Because this specification specifies the mapping for all possible values of @bit_bound, there is no need to specify the underlying type for the default as it is 32 per IDL 4.2.

Future versions of this specification or IDL RTFs may include further clarifications to facilitate the understanding of default behaviors.

The spec says for bounded strings " For interoperability purposes, implementers shall always define omg::types::bounded_string<N>, which may be declared as an alias to std::string". There is a problem when a bounded string is mapped to just a typedef/using of a std::string. At that moment it is not a distinct type so the traits will not work, as example, when defining this IDL

typedef string<5> bounded_bar1;
typedef string<5> bounded_bar2;

And both map to just std::string, the following code will not work correctly

if (traits<Test::bounded_bar>::is_bounded ()) {}
if (traits<Test::bounded_bar>::bound () == 5) {}

Clarify that omg::types::bounded_<type><N> shall always be defined for potential trait specializations

With this resolution, the specification makes it clear that implementers shall always define omg::types::bounded_string<N> for potential use of trait specializations.

Future versions of this spec may explore different avenues for mapping bounded strings.

In the example code it lists "AnException(int32_t error_code, const char* what)

", but the what argument is not described anywhere, it is not in IDL, where is it coming from?

*Add requirement for implementation of what() method in exceptions *

Specify that the mapped Exception class shall implement the virtual what() method declared by std::exception, which the mapped Exception inherits from, and at a minimum provide a mechanism to provide such explanatory information in the explicit constructor that provides a mechanism to set the value of all exception members.

1.
In table 7.1 5th row (struct is_bounded) right column:

Indicates whether a type is bounded or not.,

→ Remove comma at end.

2.
In section A.1.7.4 TypeCode Replacement, there is a small typo,

Omg::types::ref_type<CORBA::TypeCode> type() const;

The beginning should be omg:: not Omg::.

3.
Section title,
D.2.1.3.3 Bitmask Tye

→ Bitmask Type

4.
In section D.3.1.3.2 Bitsets,

IDL biset types shall be mapped as defined in Clause 7.14.3.2 of this specification.

→ bitset types

Fix typos identified in this issue

This issue addressed all reported typos.

The bit_bound annotation can also be applied to an enum, so add bit_bound/underlying_type also to the enum traits

Add bit_bound and underlying_type traits to enums

This issue adds the bit_bound and underlying_type traits to enums consistently to the traits we have added to bitmasks.

The spec should be describe that bitmasks do follow the same argument passing rules as a enum/basic type, for an in argument they should be passed by value. Also for union accessors it should describe that they follow the enum rules

No longer an issue with resolution of IDL4CPP-11

Per the resolution of IDL4CPP-11, Bitmasks no longer result on built-in types or enums, so the rules for the rest of types apply.

Can a default annotation be applied to string types (bounded and unbounded)?

@default on string and wstring are allowed

While there is no example indicating how to use @default with strings or wstrings, there's nothing in the IDL4 specification or its C++ language mapping that prevents its use. No changes in the spec are necessary.

Section 7.2.4.2 Template Types makes available types in namespace omg::types that may be used as substitutes for the native C++ types.

The same does not apply to section 7.2.4.4 for arrays.
Please change the first phrase in 7.2.4.4 to:

Add omg::types::arrays for consistency

This issue adds the definition of omg::types::arrays, for consistency with other types that are mapped to either std:: equivalents or omg::types.

The spec maps a constant within an IDL interface to a constexpr member, but this doesn't compile:

i.cpp:8:1: error: non-static data member ‘PI’ declared ‘constexpr’
8 | constexpr double PI = 3.14;

Constants declared in the context of an interface shall be marked as static

To avoid build errors, constants declared in the context of an interface (full) shall be mapped to static constexpr.

For anyone using the class mapping of a structure the spec should also provide a class mapping for a bitset, that way they have a consistent API. Also why is a inherited bitset not mapped to C++ inheritance, by removing the inheritance the inheritance relation can't be used in C++ also

Change bitset inheritance behavior

This issue replaces the inheritance behavior for bitsets, which was based on adding inherited members to the struct, to a C++ inheritance scheme.

it is not clear what the traits is_bounded/bound are for non string/sequence types, for example a long. The spec says the trait is empty, but it is there, what is the value then. When looking at the std traits, it seems better that is_bounded returns false_type (follows https://en.cppreference.com/w/cpp/types/is_bounded_array) and bound returns 0 (follows https://en.cppreference.com/w/cpp/types/rank)

Remove is_bounded and bound traits from the list of common type traits

This resolution removes is_bounded and bound traits from the list of common type traits, as they only applied to types that can may have bounds, such as strings, sequences, or maps.

Should the C++ mapping document specify that @value annotations, if specified for one field, MUST be used on all fields.

This issue shall be addressed in IDL 4.3 RTF

We have filed IDL43-93 to address this issue in the IDL 4.3 RTF. The IDL to C++ Language Mapping specification does not need to define the syntax or scope of @value at this moment.

The mapping of external in this section is not similar to 7.17.4. In 7.17.4 it is a std::shared_ptr, so a pointer to something which is shared, when copying the struct where the external is a member of the external data will not be copied, just a refcount increment, but in D2.3.4 it is descriibed that when the struct which contains the external member is copied a deep copy is done. Also what is exactly "copy functions", is that the copy constructor or something else? This looks very unspecified, the user doesn't know the exact API and the semantics should be the same as in 7.17.4

Add extra restriction to the mapping of @external

For consistency purposes, this resolution adds an extra restriction to the mapping of @external.

As far as I understand it enum_prefix/enum_postfix are a way to change the enum mapping but I don't understand why, because the enums itself are added to the enclosing module, so the following IDL is already illegal, red can't be defined twice
scope. For example the following IDL is illegal

enum Colors { red, green, blue };
enum Colors2 { red, green, blue };

See IDL4.2, chapter 7.5.2, search for "Enumeration value names are introduced into the enclosing scope and then are treated like any other declaration in that

Enum prefixes and postfixes prevent conflicts with external C++ code

In C++ 98, the mapping of IDL enums does not prevent conflicts of IDL-defined literals with the C++ code a user might be mixing or linking against. That's why even though the IDL grammar does not allow conflicts in literals within the same IDL code (such conflicts are illegal in IDL), an end user might have conflicts in the generated code, which can only be prevented adding prefix or postfix enum literal names. Mapping enums to enum classes does address this issue, but such functionality is unavailable in C++ 98.

For IDL type VT1 the example code is not correct:

• virtual is missing for all pure virtual methods
• a_long_attr is an attribute, so "int32_t& a_long_attr() = 0;" should be removed

Fix mapping of VT1 in example mapping of Clause 7.6 "Value Types"

The C++ mapping for VT1 in Clause 7.6 contains a number of errors, this resolution addresses such errors providing the correct code snippet.

The mapping says "IDL strings shall be mapped to C++ std::string, or to a type named omg::types::string that delivers std::string semantics", but when I now have CORBA and use an interface with operation "void set_string (in string text);", what is now the C++ signature I have to override in my servant implementation, it could be "void set_string(const std::string& text) override;", or "void set_string(const omg::types::string& text) override;" which could lead to portability issues when some vendor uses std::string and another omg::types::string. For defining data it doesn't make a real difference, but it is a problem when using CORBA

Question has been addressed in Jira Issue

There is nothing to change in the specification. Future versions of the specification could add further instructions on how to deal with various implementations of omg::types::strings.

The text describes that the destructor and other constructors are protected, but the example code shows them all public, the C++ code should be updated to match the text

Fix indentation of protected and replace use of std::string with omg::types::string

This issue fixes the indentation of protected and replaces the use of std::string with omg::types::string in clause A.1.10.2.

What happens when the following is present in an IDL file:

typedef @min(2) @max(1) long Huh;

Issue to be addressed in IDL 4.3 RTF

This issue refers to the IDL grammar, rather than the language mapping. Therefore, it will be addressed in the IDL 4.3 RTF.

The valuetype VT2 is defined in IDL, but the example C++ code is not correct, it says

class VT2 : public virtual VT1 {
public:
virtual void op() = 0;
virtual int32_t third_long() const = 0;
virtual int32_t& third_long() const = 0;
virtual void third_long(int16_t value) = 0;
};

But it should be

class VT2 : public virtual VT1 {
public:
virtual void op() = 0;
virtual int32_t third_long() const = 0;
virtual int32_t& third_long() = 0;
virtual void third_long(int32_t value) = 0;
};

Fix mapping of VT2 in Clause 7.6 "Value Types"

Clause 7.6 "Value Types" contains a typo in the mapping of a long, this issue fixed the typo mapping the IDL long to int32_t instead of int16_t.

Why does this example code for "C.2.1.2.4 Naming Data Types" map the IDL struct to a C++ class, that is the alternative mapping, but as far as I understand it the mapping prefers to map a IDL struct to a C++ struct by default

Fix mapping of naming data types in C.2.1.2.4

In clause C.2.1.2.4, IDL struct MyType is mapped to a class, instead of a struct, which is the correct mapping. This issue addresses that mistake.

The section is silent on maps.

Refer to Maps in clause A.1.2.4

Clause A.1.2.4, which addresses the exception type for bounded types shall refer to strings, wstrings, sequences, and maps.

A link is broken, it mentions:

Maps shall provide the following specializations for the type traits defined in Clause Error: Reference source not found:

Fix broken cross references

The specification includes a number of broken cross references. This issue replaces broken cross references with the appropriate sections.

The text:
"Inherits std::integral_constant<uint32_t, b> where b indicates
the bit_bound of the bitmask."

should instead state

Defines value - Where the value is mapped as the underlying bit_bound of the bit mask.

Type traits may be inherited

Type traits may be inherited, and it is indeed a common practice, so the mapping should remain as is. The restriction would be to inheritance of classes that have a state, such as std::vector or std::string.

As all types in the spec are lower case, shouldn't IDL fixed be mapped to C++ fixed (all lower case)

Renaming Fixed to fixed for consistency purposes

This specification defines common C++ types using lowercase names, such as bounded_string, bounded_sequence, etc. Therefore, for consistency purposes, the Fixed class shall be renamed to fixed.

The example for interfaces full lists in IDL the attribute "an_attribute", but the cod example uses "attribute", which is not correct

Fix example mapping of attribute in Interfaces - Full

The example for Interfaces - Full contains an error in the mapped name of an attribute.

In the example code there is "using Length = long;", but long is not portable, shouldn't be this be int32_t?

Replace use of C++ long in example code with int32_t

IDL longs are mapped to C++ int32_t, the example in clause 7.2.4.6 makes the mistake of mapping a typedef long to a using of long instead of int32_t.

In the example code it lists
const int16_t& a_short()

This should be a return by value, using `const int16_t` will lead to a compiler warning, so it should be 'int16_t', all accessors should be checked.

Update mapping of a_short in example code

Accessors for members of the union for which the mapped type results in primitive types and enums shall return by value. Therefore, the updated example reflects that:

// ...
void a_long(int32_t value) {...}
int32_t a_long() const {...}
// ...

The updated example also includes members that result in non-primitive types like strings or structs.

As part of the resolution of this issue we also provide a method to set the discriminator to a specific legal value when the union definition has an implicit default member (the current mapping only provides a _default() that sets the discriminator to a legal default value).

We also mandate that non-primitive members provide two getters:

const Foo& a_foo() const {}
Foo& a_foo() {}

and primitive members the following two getters:

int16_t a_long() const {}
int16_t& a_long() {}

If the getter selects a branch that does not match the discriminator value, the getter should throw an error.

In the example code the destructors in classes that derive from a base should be marked override, not virtual, this applies to

• 6.18.4: Change "virtual ~Example();" to "~Example() override;" and "virtual ~OBV_Example();" to "~OBV_Example() override;"
• 6.20: Change "virtual ~Exception();" to "~Exception() override;" and "virtual ~SystemException();" to "~SystemException() override;"
• 6.25: Change "virtual ~MyLocalIF();" to "~MyLocalIF() override;"
• 6.26.6: Change "virtual ~A_skel ();" to "~A_skel () override;" and "virtual ~A_impl ();" to "~A_impl () override;"
• 6.26.8: Change "virtual ~TIE() = default;" to "~TIE() override = default;"

Convert virtual destructors in overrides

To resolve this issue, all virtual destructors are converted to overrides.

Reporting this for bounded sequences, but very likely this has to be addressed for all bounded types. The spec says "NOTE—In any case, bound checking on bounded sequences may be performed at serialization time, raising an
exception or reporting an error if necessary", but this should be more strict to my idea, the bounds check must be performed, and the spec should specify which exception is raised when the bound is not respected

Specify that middleware implementations may mandate specific bound checking mechanisms

This issue tries to convey that middleware implementations may introduce more explicit bound checking mechanisms.

The @value annotation does not fully specify the behavior. That includes:

• The behavior for unannotated values.
• The behavior for repeated values. Is it possible like in C++?

DDS-XTYPES does specify the behavior of @value and such description should be added to IDL:

7.3.1.2.1.5 Enumerated Literal Values
Prior to this specification, it was impossible to indicate that objects of enumerated types could be stored using an integer size other than 32 bits. This specification uses the @bit_bound annotation defined in Sub Clause 8.3.4.1 of [IDL] for this purpose.

It is important to note that the value member of the annotation may take any value from 1 to 32, inclusive, when this annotation is applied to an enumerated type. Furthermore, prior to this specification, it was impossible to provide an explicit value for an enumerated literal. The value was always inferred based on the definition order of the literals. That behavior is still supported. However, additionally, this specification allows enumerated literals to be given explicit custom values, just as they can be in the C and C++ programming
languages. This can be done by means of the @value annotation defined in Sub Clause 8.3.1.5 of [IDL], which may be applied to individual literals.
It is permitted for some literals in an enumerated type to bear the @value annotation while others do not. In such cases, as in C and C++ enumerations, implicit values are assigned in a progression starting from the most-recently specified value (or an implicit value of zero for the first literal, if there is no previous specified value) and adding one with each successive literal.

Issue to be addressed in IDL4.

This issue should be addressed in the IDL 4.3 RTF.

This new spec seems to try to support anonymous types (at least it doesn't disallow them like IDL to C++11), but how are then the type traits provided for anonymous types. For example the type

struct TopicA

;

Is there a way for the user to get the IDL traits for bound/is_bounded for the l5s member? If that is not possible for anonymous types this language mapping should make that clear, that anonymous types don't have a type trait.

Add a note to clause 7.15 to illustrate type traits for anonymous types

This resolution adds an example that shows how to aquire acquire the declared type of a struct member and use it in a traits declaration.

This mapping defines the bound as std::integral_constant<size_t, b> but in most compilers size_t is 64bit where sequences/strings are bound to 32bit by IDL, so shouldn't size_t be replaced by uint32_t. If this is changed here, also check all other size_t types in this mapping

IDL Language Mappings Should Not Impose Bound Limits

Limits on bound sizes are imposed by the serialization format of middleware such as CORBA and DDS (e.g., CORBA specifies the behavior for the serialization for sequences bigger than uint32_t, and that should be enough for the middleware implementer). IDL and the corresponding IDL language mappings need not impose such limits.

The underlying_type traits is maybe something to remove, std::underlying_type does the same at the moment bitmask is implemented using an enum. Also I am doubting whether an enum or enum class is the best mapping, have you tried to implement some user code, there is no operator bool() for example, some for checking if a bit is set we always have to cast. Maybe mapping to a class with operator is easier for the user, or use an enum class with operators.

This issue will be addressed in IDL4CPP-11

The resolution of this issue will be addressed in IDL4CPP-11.

According to IDL4, annotations like range need to specify the arguments that are set. However, in the example we include for the mapping of @range we don't specify min and max in the arguments.

We need to address that as follows:

struct StructureOfRangedValues {
    @range(min=-10, max=10) long x;
};

Fix @range example

This issue fixes the use of @range in the examples.

As there is no conversion implicit from a std::string_view to a std::string there is no possibility to easily pass now a IDL defined string to an operation expecting an IDL string input, will this mapping proposal work for the users?

Middleware Implementations Should Deal with Conversions

Solving conversion issues would be a task for the specific middleware implementation. If the mapping were to use char*, implementers would have to deal with the same issue.

The spec says " In addition to the methods as described in 6.14, the class has an explicit constructor accepting values for each struct member in the order they are specified in IDL.", all values should be by value, so better it is to day " In addition to the methods as described in 6.14, the class has an explicit constructor accepting values for each struct member by value in the order they are specified in IDL."

This issue is referring to a different specification

This issue is actually pointing to the IDL to C++11 specification.

For instance is this legal:

struct S

;

typedef @min(2) long MyMin;
typedef @max(5) MyMin MyMinMax;
typedef @default(3) MyMinMax defaultedMyMinMax;

Issue Refers to IDL Grammar

This issue will be addressed in the IDL 4.3 RTF, as it refers to the IDL grammar rather than the language mapping.

Each annotation specified, either pre-defined or user-defined/custom, should discuss how it interacts with the IDL4 grammar such that the annotation has the desired effect (e.g. @range on a union).

This issue belongs in the IDL4 RTF.

This issue should be addressed in IDL4.

The spec describes that with CORBA an invalid usage of union can result in a CORBA::BAD_PARAM exception, for bounded string/vector also a CORBA::BAD_PARAM, but when not using CORBA when IDL4CSP11-9 is resolved a different exception is thrown for an union (for vector it seems this still has to be added also). When a user now has appliation code that uses IDL defined data types with and without CORBA their exception code is different. In order to achieve portability of user code I think the spec should define a omg::bad_param alias type which maps to CORBA::BAD_PARAM or the selected std exception. That way user code can be written portable when the user has data types defined by IDL and wants to support use cases with and without CORBA

Error handling and exceptions will be treated in feature versions of the spec

By design, the current version of this specification does not specify explicit mechanisms to report errors (e.g., which exception should be thrown or whether exceptions should be thrown at all). The intention was to be able to support both environments where exceptions are valid and are allowed, as well as environments where applications may not throw or handle exceptions.

Future versions of this specification may define explicit mechanisms to report errors, and will aim to harmonize them across middleware implementations.

Spec says:

If the union has a default case, the default constructor shall initialize the discriminator, and the selected
member field following the initialization rules described in Clause 7.2.4.3.1. If it does not, the default
constructor shall initialize the union to the first discriminant value specified in the IDL definition.

But when there is an implicit default member, that should be selected, for example, the example below (

  enum DataType
  {
    dtEmpty,
    dtLong,
    dtShort,
    dtString,
    dtPoint,
    dtTrack,
    dtGlobal
  };

  union Data switch (DataType)
    {
      case dtLong: long longData;
      case dtShort: short shortData;
      case dtString: string stringData;
      case dtPoint: string pointData;
      case dtTrack: string trackData;
      case dtGlobal: string globalData;
      // by default (implicit), empty union
    };

Defer the resolution of this issue to the resolution of the related IDL RTF issue

This issue is deferred until the resolution IDL43-94.

Not sure where (not necessarily in the noted paragraph) should include a reference to an electronic mapping file as an example. Could also include test sample IDL and corresponding C++ output. This should probably be informational and could also be in an Annex. The key reason is to show the actual transformation as well as a test/verification of the mappings. Perhaps also include versions of tools/compiler used in which the transformations were validated. In addition unit tests could be referenced or included as informative content to the spec.

Example Transformations To Be Incoporated in Future Versions of the Standard

We think this is an interesting concept, but it hasn't been addressed in the existing IDL4 to Language Mappings. To better align version 1.0 with the existing mappings, we think this proposal should be addressed in subsequent RTFs.

The spec in 7.4 says that only basic types/enum should be passed by value for an in argument, but that should also include all shared_ptr types, see https://herbsutter.com/2013/06/05/gotw-91-solution-smart-pointer-parameters/ for some information (more can be found on the internet).

Passing shared_ptr by value could introduce certain issues

If we pass by value we could introduce certain issues, like locking (having to increase the reference counter). It's unclear to us whether this would be beneficial. If we pass it by const reference, we're guaranteeing that the object won't be destroyed until the end of the function.

After a discussion within the IDL WG, we agreed on leaving the mapping as it currently is and reevaluate improvements to the mapping in future RTFs.

A few remarks on 7.17.3:

• formatting of code in text is not done
• what is the impact of range when applied to a typedef of a basic type on all argument passing rules when that typedef is passed with a function, stored in an union or exception?
• how should min/max be implemented, shouldn't also omg::types::ranged be used?
• what should be done when min/max but also range are used as annotation?

Defer until the resolution of the related IDL RTF issue

The IDL RTF must address some of the underlying issues when applying @min, @max, @range, and @default annotations (e.g., see IDL43-89). Therefore, we are deferring the implementation of an updated mapping until the IDL RTF tackles the underlying issue.

I find it not very maintainable to extend in this spec silently the IDL to C++11 language mapping. In the v1.7 revision I think all is already mapped, so I propose to remove D.3 completely from this spec

Defer to future versions of this specification

Future versions of this specification may remove Clause D.3 or replace its content with pointers to the appropriate clauses in the IDL to C++11 v1.7 specification. However, because this specification originally targeted v1.6 of such specification, it will require some research and agreement on the best path forward within the MARS IDL WG.

Prior to C++20, the following template cannot be instantiated:

template <typename T, const T thingy>
class MyTemplate {};

MyTemplate<double, 5.0> x;

Mapping for min, max, and ranged remains the same

In this case, an alternative implementation would introduce too much complexity. For that reason, we propose to defer this issue to future RTFs. State of the art compilers supporting C++20 work with the proposed mapping. Future RTFs may add additional information for implementers that require compatibility with earlier compilers.

Keys of maps can be any IDL type. How should the generated C++ code support key comparison of any type?

List of Valid Key Types for Maps Will Be Provided in IDL 4.3

This issue is tightly coupled with IDL43-70, which aims to determine the list of valid key types for IDL maps. Because the current IDL4 to C++ Language mapping targets IDL 4.2 (which does not specify the list of valid types), this FTF will no't be adding further restrictions.

Future versions of the IDL4 to C++ Language mapping may add key map type restrictions compliant with the resolution of IDL43-70.

the code examples use = and {} to initialize the constants, recommend to make this consistent by always using {}

Defer to future revisions of the specification

In this case, we chose to use {} only when necessary, and use = instead for when both options were possible. Having said that, given the current trend leaning towards the use of {}, future versions of this specification may change this coding convention.

With C++20 it would be nice when the mapping would specify a module mapping (see https://en.cppreference.com/w/cpp/language/modules).

C++20 Modules to be explored in future RTFs

The use of C++20 modules will be explored in future versions of this specification.