Chapter 3

Clients attempting to acquire a selection must set the time value of the SetSelectionOwner request to the timestamp of the event triggering the acquisition attempt, not to CurrentTime. A zero-length append to a property is a way to obtain a timestamp for this purpose; the timestamp is in the corresponding PropertyNotify event

If the time in the SetSelectionOwner request is in the future relative to the server's current time or is in the past relative to the last time the specified selection changed hands, the SetSelectionOwner request appears to the client to succeed, but ownership is not actually transferred.

Because clients cannot name other clients directly, the specified owner window is used to refer to the owning client in the replies to GetSelectionOwner, in SelectionRequest and SelectionClear events, and possibly as a place to put properties describing the selection in question. To discover the owner of a particular selection, a client should invoke GetSelectionOwner, which is defined as follows:

GetSelectionOwner

selection: ATOM

owner: WINDOW or None

Convention

Clients are expected to provide some visible confirmation of selection ownership. To make this feedback reliable, a client must perform a sequence like the following:

SetSelectionOwner(selection=PRIMARY, owner=Window, time=timestamp)
owner = GetSelectionOwner
(selection=PRIMARY) if (owner != Window) Failure

If the SetSelectionOwner request succeeds (not merely appears to succeed), the client that issues it is recorded by the server as being the owner of the selection for the time period starting at the specified time.

3.2.2 Responsibilities of the Selection Owner

When a requestor wants the value of a selection, the owner receives a SelectionRequest event, which is defined as follows:

Selection Request

owner: WINDOW
selection: ATOM
target: ATOM
property: ATOM or None
requestor: W1NDOW
time: TIMESTAMP or CurrentTime

The specified owner and selection will be the values that were specified in the SetSelectionOwner request. The owner should compare the timestamp with the period it has owned the selection and, if the time is outside, refuse the SelectionRequest by sending the requestor window a SelectionNotify event with the property set to None (by means of a SendEvent request with an empty event mask).

More advanced selection owners are free to maintain a history of the value of the selection and to respond to requests for the value of the selection during periods they owned it even though they do not own it now.

If the specified property is None, the requestor is an obsolete client. Owners are encouraged to support these clients by using the specified target atom as the property name to be used for the reply.

Otherwise, the owner should use the target to decide the form into which the selection should be converted. Some targets may be defined such that requestors can pass parameters along with the request. The owner will find these parameters in the property named in the selection request. The type, format, and contents of this property are dependent upon the definition of the target. If the target is not defined to have parameters, the owner should ignore the property if it is present. If the selection cannot be converted into a form based on the target (and parameters, if any), the owner should refuse the SelectionRequest as previously described.

If the specified property is not None, the owner should place the data resulting from converting the selection into the specified property on the requestor window and should set the property's type to some appropriate value, which need not be the same as the specified target.

Convention

All properties used to reply to SelectionRequest events must be placed on the requestor window.

In either case, if the data comprising the selection cannot be stored on the requestor window (for example, because the server cannot provide sufficient memory), the owner must refuse the SelectionRequest, as previously described. See also Section 3.2.5.

If the property is successfully stored, the owner should acknowledge the successful conversion by sending the requestor window a SelectionNotify event (by means of a SendEvent request with an empty mask). SelectionNotify is defined as follows:

SelectionNotify

requestor: WINDOW
selection, target: ATOM
property: ATOM or None
time: TIMESTAMP or CurrentTime

The owner should set the specified selection, target, time, and property arguments to the values received in the SelectionRequest event. (Note that setting the property argument to None indicates that the conversion requested could not be made.)

Convention

The selection, target, time, and property arguments in the SelectionNotify event should be set to the values received in the SelectionRequest event.

If the owner receives more than one SelectionRequest event with the same requestor, selection, target, and timestamp, it must respond to them in the same order in which they were received.

Rationale

It is possible for a requestor to have multiple outstanding requests that use the same requestor window, selection, target, and timestamp, and that differ only in the property. If this occurs, and one of the conversion requests fails, the resulting SelectionNotify event will have its property argument set to None. This may make it impossible for the requestor to determine which conversion request had failed, unless the requests are responded to in order.

The data stored in the property must eventually be deleted. A convention is needed to assign the responsibility for doing so.

Convention

Selection requestors are responsible for deleting properties whose names they receive in SelectionNotify events (see Section 3.2.4) or in properties with type MULTIPLE.

A selection owner will often need confirmation that the data comprising the selection has actually been transferred. (For example, if the operation has side effects on the owner's internal data structures, these should not take place until the requestor has indicated that it has successfully received the data.) Owners should express interest in PropertyNotify events for the specified requestor window and wait until the property in the SelectionNotify even has been deleted before assuming that the selection data has been transferred.
For the MULTIPLE request, if the different conversions require separate confirmation, the selection owner can also watch for the deletion of the individual properties named in the property in the SelectionNotify event.
When some other client acquires a selection, the previous owner receives a SelectionClear event, which is defined as follows:

SelectionClear

owner: WINDOW
selection: ATOM
time: TIMESTAMP

The timestamp argument is the time at which the ownership changed hands, and the owner argument is the window the previous owner specified in its SetSelectionOwner request.

If an owner loses ownership while it has a transfer in progress (that is, before it receives notification that the requestor has received all the data), it must continue to service the ongoing transfer until it is complete.

If the selection value completely changes, but the owner happens to be the same client (for example, selecting a totally different piece of text in the same xterm as before), then the client should reacquire the selection ownership as if it were not the owner, providing a new timestamp. If the selection value is modified, but can still reasonably be viewed as the same selected object, the owner should take no action. (The division between these two cases is a matter of judgment on the part of the software developer)

3.2.3 Giving Up Selection Ownership

Clients may either give up selection ownership voluntarily or lose it forcibly as the result of some other client's actions.

3.2.3.1 Voluntarily Giving Up Selection Ownership

To relinquish ownership of a selection voluntarily, a client should execute a SetSelectionOwner request for that selection atom, with owner specified as None and the time specified as the timestamp that was used to acquire the selection.

Alternatively, the client may destroy the window used as the owner value of the SetSelectionOwner request, or the client may terminate. In both cases, the ownership of the selection involved will revert to None.

3.2.3.2 Forcibly Giving Up Selection Ownership

If a client gives up ownership of a selection or if some other client executes a SetSelectionOwner for it and thus reassigns it forcibly, the previous owner will receive a Selection Clear event. For the definition of a Selection Clear event, see Section 3.2.2.

The timestamp is the time the selection changed hands. The specified owner is the window that was specified by the current owner in its SetSelectionOwner request.

3.2.4 Requesting a Selection

A client that wishes to obtain the value of a selection in a particular form (the requestor) issues a ConvertSelection request, which is defined as follows:

ConvertSelection

selection, target: ATOM
property: ATOM or None
requestor: WINDOW
time: TIMESTAMP or CurrentTime

The selection argument specifies the particular selection involved, and the target argument specifies the required form of the information. For information about the choice of suitable atoms to use, see Section 3.2.6. The requestor should set the requestor argument to a window that it created; the owner will place the reply property there. The requestor should set the time argument to the timestamp on the event that triggered the request for the selection value. Note that clients should not specify CurrentTime.

Convention

Clients should not use CurrentTime for the time argument of a ConvertSelection request. Instead, they should use the timestamp of the event that caused the request to be made.

The requestor should set the property argument to the name of a property that the owner can use to report the value of the selection. Requestors should ensure that the named property does not exist on the window before issuing the ConvertSelection request. The exception to this rule is when the requestor intends to pass parameter with the request.

(This requirement is new in vermin 2.0, and in general, existing clients do not conform to this requirement To prevent these clients from breaking, no existing targets should be extended to take parameters until sufficient time has passed for clients to be updated Note that the MULTIPLE target was defined to take parameters in version 1 0 and its definition is not changing There is thus no conformance problem with MULTIPLE.)

Rationale

It is necessary for requestors to delete the property before issuing the request so that the target can later be extended to take parameters without introducing an incompatibility. Also note that the requestor of a selection need not know the client that owns the selection nor the window on which the selection was acquired.

Some targets may be defined such that requesters can pass parameters along with the request. If the requestor wishes to provided parameters to a request, they should be placed in the specified property on the requestor window before the requestor issues the ConvertSelection request, and this property should be named in the request.

Some targets may be defined so that parameters are optional. If no parameters are to be supplied with the request of such a target, the requestor must ensure that the property does not exist before issuing the ConvertSelection request.

The protocol allows the property field to be set to None, in which case the owner is supposed to choose a property name. However, it is difficult for the owner to make this choice safely.

Conventions

1. Requestors should not use None for the property argument of a ConvertSelection request.
2. Owners receiving ConvertSelection requests with a property argument of None are talking to an obsolete client. They should choose the target atom as the property name to be used for the reply.

The result of the ConvertSelection request is that a SelectionNotify event will be received. For the definition of a SelectionNotify event, see Section 3.2.2.

The requestor, selection, time, and target arguments will be the same as those on the ConvertSelection request.

If the property argument is None, the conversion has been refused. This can mean either that there is no owner for the selection, that the owner does not support the conversion implied by the target, or that the server did not have sufficient space to accommodate the data.

If the property argument is not None, then that property will exist on the requestor window. The value of the selection can be retrieved from this property by using the GetProperty request, which is defined as follows:

GetProperty

window: WINDOW
property: ATOM
type: ATOM or AnyPropertyType
long-offset, long-length: CARD32
delete: BOOL

type: ATOM or None
format: {0, 8,16, 32}
bytes-after: CARD32
value: LISTofINT8 or LISToflNT16 or LISTofINT32

When using GetProperty to retrieve the value of a selection, the property argument should be set to the corresponding value in the SelectionNotify event. Because the requestor has no way of knowing beforehand what type the selection owner will use, the type argument should be set to AnyPropertyType. Several GetProperty requests may be needed to retrieve all the data in the selection; each should set the long-offset argument to the amount of data received so far, and the size argument to some reasonable buffer size (see Section 3.2.5). If the returned value of bytes-after is zero, the whole property has been transferred.

Once all the data in the selection has been retrieved (which may require getting the values of several properties-see Section 3.2.7), the requestor should delete the property in the SelectionNotify request by using a GetProperty request with the delete argument set to True. As previously discussed, the owner has no way of knowing when the data has been transferred to the requestor unless the property is removed.

Convention

The requestor must delete the property named in the SelectionNotify once all the data has been retrieved. The requestor should invoke either DeleteProperty or GetProperty (delete==True) after it has successfully retrieved all the data in the selection. For further information, see Section 3.2.5.

3.2.5 Large Data Transfers

Selections can get large, which poses two problems:

• Transferring large amounts of data to the server is expensive.



• All servers will have limits on the amount of data that can be stored in properties. Exceeding this limit will result in an Alloc error on the ChangeProperty request that the selection owner uses to store the data.

The problem of limited server resources is addressed by the following conventions:

Selection owners should transfer the data describing a large selection (relative to the maximum-request-size they received in the connection handshake) using the INCR property mechanism (see Section .3.2.7.2).

Any client using SetSelectionOwner to acquire selection ownership should arrange to process Alloc errors in property change requests. For clients using Xlib, this involves using the XSetErrorHandler function to override the default handler.

A selection owner must confirm that no Alloc error occurred while storing the properties for a selection before replying with a confirming SelectionNotif event.

When storing large amounts of data (relative to maximum-request-size), clients should use a sequence ofChangeProperty(mode=Append) requests for reasonable quantities of data. This avoids locking servers up and limits the waste of data an Alloc error would cause.

If an Alloc error occurs during the storing of the selection data, all properties stored for this selection should be deleted and the ConvertSelection request should be refused (see Section 3.2.2).

To avoid locking servers up for inordinate lengths of time, requestors retrieving large quantities of data from a property should perform a series of GetProperty requests, each asking for a reasonable amount of data.

Single-threaded servers should take care to avoid locking up during large data transfers.

3.2.6 Use of Selection Atoms

Defining a new atom consumes resources in the server that are not released until the server reinitializes. Thus, reducing the need for newly minted atoms is an important goal for the use of the selection atoms.

3.2.6.1 Selection Atoms

There can be an arbitrary number of selections, each named by an atom. To conform with the inter-client conventions, however, clients need deal with only these three selections:

• PRIMARY
• SECONDARY
• CLIPBOARD

Other selections may be used freely for private communication among related groups of clients.

3.2.6.1.1 The PRIMARY Selection

The selection named by the atom PRIMARY is used for all commands that take only a single argument and is the principal means of communication between clients that use the selection mechanism.

3.2.6.1.2 The SECONDARY Selection

The selection named by the atom SECONDARY is used:

• As the second argument to commands taking two arguments (for example, "exchange primary and secondary selections")
• As a means of obtaining data when there is a primary selection and the user does not want to disturb it

3.2.6.1.3 The CLIPBOARD Selection

The selection named by the atom CLIPBOARD is used to hold data that is being transferred between clients, that is, data that usually is being cut or copied, and then pasted. Whenever a client wants to transfer data to the clipboard:

• It should assert ownership of the CLIPBOARD.
• If it succeeds in acquiring ownership, it should be prepared to respond to a request for the contents of the CLIPBOARD in the usual way (retaining the data to be able to return it).
  The request may be generated by the clipboard client described below.
• If it fails to acquire ownership, a cutting client should not actually perform the cut or provide feedback that would suggest that it has actually transferred data to the clipboard.
  

The owner should repeat this process whenever the data to be transferred would change.

Clients wanting to paste data from the clipboard should request the contents of the CLIPBOARD selection in the usual way.

Except while a client is actually deleting or copying data, the owner of the CLIPBOARD selection may be a single, special client implemented for the purpose. This client maintains the content of the clipboard up-to-date and responds to requests for data from the clipboard as follows:

• It should assert ownership of the CLIPBOARD selection and reassert it any time the clipboard data changes.
• If it loses the selection (because another client has some new data for the clipboard), it should:

  -	Obtain the contents of the selection from the new owner by using the timestamp in the SelectionClear event.
  -	Attempt to reassert ownership of the CLIPBOARD selection by using the same timestamp.
  -	Restart the process using a newly acquired timestamp if this attempt fails. This timestamp should be obtained by asking the current owner of the CLIPBOARD selection to convert it to a TIMESTAMP. If this conversion is refused or if the same timestamp is received twice, the clipboard client should acquire a fresh timestamp in the usual way (for example by a zero-length append to a property).

• It should respond to requests for the CLIPBOARD contents in the usual way.

A special CLIPBOARD client is not necessary. The protocol used by the cutting client and the pasting client is the same whether the CLIPBOARD client is running or not. The reasons for running the special client include:

• Stability - If the cutting client were to crash or terminate, the clipboard value would still be available.
• Feedback - The clipboard client can display the contents of the clipboard.
• Simplicity - A client deleting data does not have to retain it for so long, thus reducing the chance of race conditions causing problems.

The reasons not to run the clipboard client include:

• Performance - Data is only transferred if it is actually required (that is, when some client actually wants the data).
• Flexibility - The clipboard data may be available as more than one target.

3.2.6.2 Target Atoms

The atom that a requestor supplies as the target of a ConvertSelection request determines the form of the data supplied. The set of such atoms is extensible, but a generally accepted base set of target atoms is needed. As a starting point for this, the following table contains those that have been suggested so far.

5) Earlier versions of this document erroneously specified that conversion of the PIXMAP target return a property of type DRAWABLE instead of PIXMAP. Implementers should be aware of this and may want to support the DRAWABLE type as well to allow for compatibility with older clients.


6) The targetsENCAPSULATED_POSTSCRIPT and ENCAPSULATED_POSTSCRIPT_INTERCHANGE are equivalent to the targets _ADOBE_EPS and _ADOBE_EPSI (respectively) that appear in the selection targets registry. The _ADOBE_targets are deprecated, but clients are encouraged to continue to support them for backward compatibility.


7) This definition is ambiguous, as the selection may be converted into any of several targets which may return differing amounts of data, The requestor has no way of knowing which, if any. of these targets corresponds to the result of LENGTH. Clients are advised that no guarantees can be made about the result of a conversion to LENGTH; its use is thus deprecated.

References (print-only documents):

1. Adobe Systems, incorporated. Portable Document Format Reference Manual. AddisonWesley, ISBN 0-201-62628-4.
2. Apple Computer, Incorporated. Inside Macintosh, Volume V. Chapter 4, "Color Quick-Draw," Color Picture Format. ISBN0-201-17719-6.
3. Adobe Systems, Incorporated. PostScript Language Reference Manual. Addison-Wesley, ISBN 0-20l - 18127-4.

It is expected that this table will grow over time.

Selection owners are required to support the following targets. All other targets are optional.

• TARGETS - The owner should return a list of atoms that represent the targets for which an attempt to convert the current selection will succeed (barring unforseeable problems-such as Alloc errors). This list should include all the required atoms.
• MULTIPLE - The MULTIPLE target atom is valid only when a property is specified on the ConvertSelection request. If the property argument in the SelectionRequest event is None and the target is MULTIPLE, it should be refused.

  When a selection owner receives a SelectionRequest(target=MULTIPLE) request, the contents of the property named in the request will be a list of atom pairs: the first atom naming a target and the second naming a property (None is not valid here). The effect should be as if the owner had received a sequence of SelectionRequest events (one for each atom pair) except that:

  -	The owner should reply with a SelectionNotify only when all the requested conversions have been performed.
  -	If the owner fails to convert the target used by an atom in the MULTIPLE property, it should replace that atom in the property with None.

  Convention

  The entries in a MULTIPLE property must be processed in the order they appear in the property. For further information, see Section 3.2.6.3.

  The requestor should delete each individual property when it has copied the data from that conversion, and the property specified in the MULTIPLE request when it has copied all the data.

  The requests are otherwise to be processed independently, and they should succeed or fail independently. The MULTIPLE target is an optimization that reduces the amount of protocol traffic between the owner and the requestor; it is not a transaction mechanism. For example, a client may issue a MULTIPLE request with two target: a data target and the DELETE target. The DELETE target will still be processed even if the conversion of the data target fails.

• TIMESTAMP - To avoid some race conditions, it is important that requestors be able to discover the timestamp the owner used to acquire ownership. Until and unless the protocol is changed so that a GetSelectionOwner request returns the timestamp used to acquire ownership, selection owners must support conversion to TIMESTAMP, returning the timestamp they used to obtain the selection.

3.2.6.3 Selection Targets with Side Effects

Some targets (for example, DELETE) have side effects. To render these targets unambiguous, the entries in a MULTIPLE property must be processed in the order that they appear in the property.

In general, targets with side effects will return no information, that is, they will return a zerolength property of type NULL. (Type NULL means the result of InternAtom on the string "NULL", not the value zero.) In all cases, the requested side effect must be performed before the conversion is accepted. If the requested side effect cannot be performed, the corresponding conversion request must be refused.

Conventions

1. Targets with side effects should return no information (that is, they should have a zero-length property of type NULL).
2. The side effect of a target must be performed before the conversion is accepted.
3. If the side effect of a target cannot be performed, the corresponding conversion request must be refused.

Problem

The need to delay responding to the ConvertSelection request until a further conversion has succeeded poses problems for the Intrinsics interface that need to be addressed.

These side effect targets are used to implement operations such as "exchange PRIMARY and SECONDARY selections."

3.2.6.3.1 DELETE

When the owner of a selection receives a request to convert it to DELETE, it should delete the corresponding selection (whatever doing so means for its internal data structures) and return a zero-length property of type NULL if the deletion was successful.

3.2.6.3.2 INSERT SELECTION

When the owner of a selection receives a request to convert it to INSERT_SELECTION, the property named will be of type ATOM_PAIR. The first atom will name a selection, and the second will name a target. The owner should use the selection mechanism to convert the named selection into the named target and should insert it at the location of the selection for which it got the INSERT_SELECTION request (whatever doing so means for its internal data structures).

3.2.6.3.3 INSERT PROPERTY

When the owner of a selection receives a request to convert it to INSERT_PROPERTY, it should insert the property named in the request at the location of the selection for which it got the INSERT_SELECTION request (whatever doing so means for its internal data structures).

3.2.7 Use of Selection Properties

The names of the properties used in selection data transfer are chosen by the requestor. The use of None property fields in ConvertSelection requests (which request the selection owner to choose a name) is not permitted by these conventions.

The selection owner always chooses the type of the property in the selection data transfer. Some types have special semantics assigned by convention, and these are reviewed in the following sections.

In all cases, a request for conversion to a target should return either a property of one of the types listed in the previous table for that property or a property of type INCR and then a property of one of the listed types.

Certain selection properties may contain resource IDs. The selection owner should ensure that the resource is not destroyed and that its contents are not changed until after the selection transfer is complete. Requestors that rely on the existence or on the proper contents of a resource must operate on the resource (for example, by copying the contents of a pixmap) before deleting the selection property.

The selection owner will return a list of zero or more items of the type indicated by the property type. In general, the number of items in the list will correspond to the number of disjoint parts of the selection. Some targets (for example, side-effect targets) will be of length zero irrespective of the number of disjoint selection parts. In the case of fixed-size items, the requestor may determine the number of items by the property size. For variable-length items such as text, the separators are listed in the following table:

It is expected that this table will grow over time.

3.2.7.1 TEXT Properties

In general, the encoding for the characters in a text string property is specified by its type. It is highly desirable for there to be a simple, invertible mapping between string property types and any character set names embedded within font names in any font naming standard adopted by the Consortium.

The atom TEXT is a polymorphic target. Requesting conversion into TEXT will convert into whatever encoding is convenient for the owner. The encoding chosen will be indicated by the type of the property returned. TEXT is not defined as a type; it will never be the returned type from a selection conversion request.

If the requestor wants the owner to return the contents of the selection in a specific encoding, it should request conversion into the name of that encoding.

In the table in Section 3.2.6.2, the word TEXT (in the Type column) is used to indicate one of the registered encoding names. The type would not actually be TEXT; it would be STRING or some other ATOM naming the encoding chosen by the owner.

STRING as a type or a target specifies the ISO Latin-l character set plus the control characters TAB (octal 11) and NEWLINE (octal 12). The spacing interpretation of TAB is context dependent. Other ASCII control characters are explicitly not included in STRING at the present time.

COMPOUND_TEXT as a type or a target specifies the Compound Text interchange format; see the Compound Text Encoding.

There are some text objects where the source or intended user, as the case may be, does not have a specific character set for the text, but instead merely requires a zero-terminated sequence of bytes with no other restriction; no element of the selection mechanism may assume that nay byte value is forbidden or that any two differing sequences are equivalent⁸⁾. For these objects, the type C_STRING should be used.

Rationale

An example of the need for C_STRING is to transmit the names of files; many operating systems do not interpret filenames as having a character set. For example, the same character string uses a different sequence of bytes in ASCII and EBCDIC, and so most operating systems see these as different filenames, and offer no way to treat them as the same. Thus no character-set based property type is suitable.

Type STRING and COMPOUND_TEXT properties will consist of a list of elements separated by null characters; other encodings will need to specify an appropriate list format.

8) Note that this different from STRING. where many byte values are forbidden, and from COMPOUND TEXT. where, for example inserting the sequence 27, 40.66 (designate ASCII into GL) at the start does not alter the meaning.

3.2.7.2 INCR Properties

Requestors may receive a property of type INCR⁹⁾ in response to any target that results in selection data;. This indicates that the owner will send the actual data incrementally. The contents of the INCR property will be an integer, which represents a lower bound on the number of bytes of data in the selection. The requestor and the selection owner transfer the data in the selection in the following manner.

The selection requestor starts the transfer process by deleting the (type==INCR) property forming the reply to the selection.

The selection owner then:

• Appends the data in suitable-size chunks to the same property on the same window as the selection reply with a type corresponding to the actual type of the converted selection. The size should be less than the maximum-request-size in the connection handshake.
• Waits between each append for a PropertyNotify(state==Deleted) event that shows that the requestor has read the data. The reason for doing this is to limit the consumption of space in the server.
• Waits (after the entire data has been transferred to the server) until a PropertyNotify(state==Deleted) event that shows that the data has been read by the requestor and then writes zero-length data to the property.

The selection requestor:

• Waits for the SelectionNotify event.
• Loops:

  -	Retrieving data using GetProperty with the delete argument True.
  -	Waiting for a PropertyNotify with the state argument NewValue.

• Waits until the property named by the PropertyNotify event is zero-length.
• Deletes the zero-length property.

The type of the converted selection is the type of the first partial property. The remaining partial properties must have the same type.

9) These properties were called INCREMENTAL in an earlier draft. The protocol for using them has changed. and so the name has changed to avoid confusion

3.2.7.3 DRAWABLE Properties

Requestors may receive properties of type PIXMAP, BITMAP, DRAWABLE, or WINDOW, which contain an appropriate ID. While information about these drawables is available from the server by means of the GetGeometry request, the following items are not:

• Foreground pixel
• Background pixel
• Colormap ID

In general, requestors converting into targets whose returned type in the table in Section 3.2.6.2 is one of the DRAWABLE types should expect to convert also into the following targets (using the MULTIPLE mechanism):

• FOREGROUND returns a PIXEL value.
• BACKGROUND returns a PIXEL value.
• COLORMAP returns a colormap ID.

3.2.7.4 SPAN Properties

Properties with type SPAN contain a list of cardinal-pairs with the length of the cardinals determined by the format. The first specifies the starting position, and the second specifies the ending position plus one. The base is zero. If they are the same, the span is zero-length and is before the specified position. The units are implied by the target atom, such as LINE_NUMBER or CHARACTER_POSITION.

3.2.8 Manager Selections

Certain clients, often called managers, take on responsibility for managing shared resources. A client that manages a shared resource should take ownership of an appropriate selection, named using the conventions described in Sections 3.1.2.3 and 3.1.2.6. A client that manages multiple shared resources (or groups of resources) should take ownership of a selection for each one.

The manager may support conversion of various targets for that selection. Managers are encouraged to use this technique as the primary means by which clients interact with the managed resource. Note that the conventions for interacting with the window manager predate this section; as a result many interactions with the window manager use other techniques.

Before a manager takes ownership of a manager selection, it should use the GetSelectionOwner request to check whether the selection is already owned by another client, and where appropriate, it should ask the user if the new manager should replace the old one. If so, it may then take ownership of the selection. Managers should acquire the selection using a window created expressly for this purpose. Managers must conform to the rules for selection owners described in Sections 3.2.1 and 3.2.2, and they must also support the required targets listed in Section 3.2.6.2.

If a manager loses ownership of a manager selection, this means that a new manager is taking over its responsibilities. The old manager must release all resources it has managed, and must then destroy the window that owned the selection. For example, a window manager losing ownership of WM_S2 must be select from Substructure Redirect on the root window of screen 2 before destroying the window that owned WM_S2.

When the new manager notices that the window owning the selection has been destroyed, it knows that it can successfully proceed to control the resource it is planning to manage. If the old manager does not destroy the window within a reasonable time, the new manager should check with the user before destroying the window itself or killing the old manager.

If a manager wants to give up, on its own, management of a shared resource controlled by a selection, it must do so by releasing the resources it is managing, and then by destroying the window that owns the selection. It should not first disown the selection, since this introduces a race condition.

Clients who are interesting in knowing when the owner of a manager selection is no longer managing the corresponding shared resource should select for StructureNotify on the window owning the selection so they can be notified when the window is destroyed. Clients are warned that after doing a GetSelectionOwner and selecting for StructureNotify, they should do a GetSelectionOwner again to ensure that the owner did not change after initially getting the selection owner and before selection for StructureNotify.

Immediately after a manager successfully acquires ownership of a manager selection, it should announce its arrival by sending a ClientMessage event. This event should be sent using the SendEvent protocol request with the following arguments:

10) We use the notation data[n] to indicate the n^th element of the LlST of INT8, LIS of INT16, or LlSToflNT32 in the data field of the ClientMessage, according to the format field. The list is indexed from zero.

Clients that wish to know when a specific manager has started should select for StructureNotify on the appropriate root window, and should watch for the appropriate MANAGER Client-Message.

3.3 COMMUNICATION BY MEANS OF CUT BUFFERS

The cut buffer mechanism is much simpler but much less powerful than the selection mechanism. The selection mechanism is active in that it provides a link between the owner and requestor clients. The cut buffer mechanism is passive; an owner places data in a cut buffer from which a requestor retrieves the data at some later time.

The cut buffers consist of eight properties on the root of screen zero, named by the predefined atoms CUT_BUFFER to CUT_BUFFERI. These properties must, at present, have type STRING and format 8. A client that uses the cut buffer mechanism must initially ensure that all eight properties exist by using ChangeProperty requests to append zero-length data to each.

A client that stores data in the cut buffers (an owner) first must rotate the ring of buffers by plus 1 by using RotateProperties requests to rename each buffer; that is, CUT_BUFFER0 CUT_B JFFER1, CUT_BUFFER1 to CUT_BUFFER2,..., and CUT_BUFFER7 to CUT_B JFFER0. It then must store the data into CUT_BUFFER0 by using a ChangeProperty request n mode Replace.

A client that obtains data from the cut buffers should use a GetProperty request to retrieve the contents of CUT_BUFFER0.

In response to a specific user request, a client may rotate the cut buffers by minus 1 by using RotateProperties requests to rename each buffer; that is, CUT_BUFFER7 to CUT_BUFFER6, CUT_BUFFER6 to CUT_BUFFER5,..., and CUT_BUFFER0 to CUT_BUFFER7.

Data should be stored to the cut buffers and the ring rotated only when requested by explicit user action. Users depend on their mental model of cut buffer operation and need to be able to identify operations that transfer data to and from.

3.4 CLIENT TO WINDOW MANAGER COMMUNICATION

To permit window managers to perform their role of mediating the competing demands for resources such as screen space, the clients being managed must adhere to certain conventions and must expect the window managers to do likewise. These conventions are covered here from the client's point of view.

In general, these conventions are somewhat complex and will undoubtedly change as new window management paradigms are developed. Thus, there is a strong bias toward defining only those conventions that are essential and that apply generally to all window management paradigms. Clients designed to run with a particular window manager can easily define private protocols to add to these conventions, but they must be aware that their users may decide to run some other window manager no matter how much the designers of the private protocol are convinced that they have seen the "one true light" of user interfaces.

It is a principle of these conventions that a general client should neither know nor care which window manager is running or, indeed, if one is running at all. The conventions do not support all client functions without a window manager running; for example, the concept of Iconic is not directly supported by clients. If no window manager is running, the concept of Iconic does not apply. A goal of the conventions is to make it possible to kill and restart window managers without loss of functionality.

Each window manager will implement a particular window management policy; the choice of an appropriate window management policy for the user's circumstances is not one for an individual client to make but will be made by the user or the user's system administrator. This does not exclude the possibility of writing clients that use a private protocol to restrict themselves to operating only under a specific window manager. Rather, it merely ensures that no claim of general utility is made for such programs.

For example, the claim is often made: The client I'm writing is important, and it needs to be on top." Perhaps it is important when it is being run in earnest, and it should then be run under the control of a window manager that recognizes "important" windows through some private protocol and ensures that they are on top. However, imagine, for example, that the " important" client is being debugged. Then, ensuring that it is always on top is no longer the appropriate window management policy, and it should be run under a window manager that allows other windows (for example, the debugger) to appear on top.

3.4.1 Client's Actions

In general, the object of the X Version 11 design is that clients should, as far as possible, do exactly what they would do in the absence of a window manager, except for the following:

• Hinting to the window manager about the resources they would like to obtain
• Cooperating with the window manager by accepting the resources they are allocated even if they are not those requested
• Being prepared for resource allocations to change at any time

3.4.1.1 Creating a Top-Level Window

A client's top-level window is a window whose override-redirect attribute is False. It must either be a child of a root window, or it must have been a child of a root window immediately prior to having been reparented by the window manager. If the client reparents the window away from the root, the window is no longer a top-level window; but it can become a top-level window again if the client reparents it back to the root.

A client usually would expect to create its top-level windows as children of one or more of the root windows by using some boilerplate like the following:

win = XCreateSimpleWindow(dpy, DefaultRootWindow(dpy), xsh.x, xsh.y, xsh.width, xsh.height, bw, bd, bg);

If a particular one of the root windows was required, however, it could use something like the following:

win = XCreateSimpleWindow(dpy, RootWindow(dpy, screen), xsh.x, xsh.y, xsh.width, xsh.height, bw, bd, bg);

Ideally, it should be possible to override the choice of a root window and allow clients (including window managers) to treat a nonroot window as a pseudo-root. This would allow, for example, the testing of window managers and the use of application-specific window managers to control the subwindows owned by the members of a related suite of clients. Doing so properly requires an extension, the design of which is under study.

From the client's point of view, the window manager will regard its top-level window as being in one of three states:

• Normal
• Iconic
• Withdrawn

Newly created windows start in the Withdrawn state. Transitions between states happen when the top-level window is mapped and unmapped and when the window manager receives certain messages. For further details, see Sections 3.4.1.2.4 and 3.4.1.4.

3.4.1.2 Client Properties

Once the client has one or more top-level windows, it should place properties on those windows to inform the window manager of the behavior that the client desires. Window managers will assume values they find convenient for any of these properties that are not supplied; clients that depend on particular values must explicitly supply them. The window manager will not change properties written by the client.

The window manager will examine the contents of these properties when the window makes the transition from the Withdrawn state and will monitor some properties for changes while the window is in the Iconic or Normal state. When the client changes one of these properties, it must use Replace mode to overwrite the entire property with new data; the window manager will retain no memory of the old value of the property. All fields of the property must be set to suitable values in a single Replace mode ChangeProperty request. This ensures that the full contents of the property will be available to a new window manager if the existing one crashes, if it is shut down and restarted, or if the session needs to be shut down and restarted by the session manager.

Convention

Clients writing or rewriting window manager properties must ensure that the entire content of each property remains valid at all times.

If these properties are longer than expected, clients should ignore the remainder of the property. Extending these properties is reserved to the X Consortium; private extensions to them are forbidden. Private additional communication between clients and window managers should take place using separate properties.

The only exception to this rule is the WM_PROTOCOLS property, which may be of arbitrary length and which may contain atoms representing private protocols; see Section 3.4.1.2.7.

The next sections describe each of the properties the clients need to set, in turn. They are summarized in the table in Section 3.4.4.

3.4.1.2.1 WM_NAMEProperty

The WM_NAME property is an uninterpreted string that the client wants the window manager to display in association with the window (for example, in a window headline bar).

The encoding used for this string (and all other uninterpreted string properties) is implied by the type of the property. The type atoms to be used for this purpose are described in Section 3.2.7.1.

Window managers are expected to make an effort to display this information. Simply ignoring WM_NAME is not acceptable behavior. Clients can assume that at least the first part of this string is visible to the user and that if the information is not visible to the user, it is because the user has taken an explicit action to make it invisible.

On the other hand, there is no guarantee that the user can see the WM_NAME string even if the window manager supports window headlines. The user may have placed the headline off-screen or have covered it by other windows. WM_NAME should not be used for application-critical information or to announce asynchronous changes of an application's state that require timely user response. The expected uses are to permit the user to identify one of a number of instances of the same client and to provide the user with noncritical state information

Even window managers that support headline bars will place some limit on the length of the WM_NAME string that can be visible; brevity here will pay dividends.

3.4.1.2.2 WM ICON NAME Property

The WM_ICON_NAME property is an uninterpreted string that the client wants to be displayed in association with the window when it is iconified (for example, in an icon label). In other respects, including the type, it is similar to WM_NAME. For obvious geometric reasons, fewer characters will normally be visible in WM_ICON_NAME than WM_NAME;

Clients should not attempt to display this string in their icon pixmaps or windows; rather, they should rely on the window manager to do so.

3.4.1.2.3 WM NORMAL HINTS Property

The type of the WM_NORMAL_HINTS property is WM_SIZE_HINTS. Its contents are as follows:

The WM_SIZE_HINTS.flags bit definitions are as follows:

To indicate that the size and position of the window (when mapped from the Withdrawn state) was specified by the user, the client should set the USPosition and USSize flags, which allow a window manager to know that the user specifically asked where the window should be placed or how the window should be sized and that further interaction is superfluous. To indicate that it was specified by the client without any user involvement, the client should set PPosition and PSize.

The site specifiers refer to the width and height of the client's window excluding borders.

The win_gravity may be any of the values specified for WINGRAVITY in the core protocol except for Unmap: NorthWest (1), North (2), NorthEast (3), West (4), Center (5), East (6), SouthWest (7), South (8), and South East (9). It specifies how and whether the client window wants to be shifted to make room for the window manager frame.

If the win_gravity is Static, the window manager frame is positioned so that the inside border of the client window inside the frame is in the same position on the screen as it was when the client requested the transition from Withdrawn state. Other values of win_gravity specify a window reference point. For NorthWest, NorthEast, SouthWest, and SouthEast the reference point is the specified outer corner of the window (on the outside border edge). For North, South, East, and West the reference point is the center of the specified other edge of the window border. For Center the reference point is the center of the window. The reference point of the window manager frame is placed at the location on the screen where the reference point of the client window was when the client requested the transition from Withdrawn state.

The min_width and min_height elements specify the minimum size that the window can be for the client to be useful. The max_width and max_height elements specify the maximum size. The base_width and base_height elements in conjunction with width_inc and height_inc define an arithmetic progression of preferred window widths and heights for nonnegative integers i and j:

width = base_width + ( i * width_inc)
height = base_height + ( j * height_inc )

Window managers are encouraged to use i and j instead of width and height in reporting window sizes to users. If a base size is not provided, the minimum size is to be used in its place and vice versa

The min_aspect and max_aspect fields are fractions with the numerator first and the denominator second, and they allow a client to specify the range of aspect ratios it prefers.

Window managers that honor aspect ratios should take into account the base size in determining the preferred window size. If a base size is provided along with the aspect ratio fields, the base size should be subtracted from the window size prior to checking that the aspect ratio falls in range. If a base size is not provided, nothing should be subtracted from the window size. (The minimum size is not to be used in place of the base size for this purpose.)

3.4.1.2.4 WM HINTS Property

The WM_HINTS property (whose type is WM_HINTS) is used to communicate to the window manager. It conveys the information the window manager needs other than the window geometry, which is available from the window itself; the constraints on that geometry, which is available from the WM_NORMAL_HINTS structure; and various strings, which need separate properties, such as WM_NAME. The contents of the properties are as follows:

The WM_HINTS.flags bit definitions are as follows:

Window managers are free to assume convenient values for all fields of the WM_HINTS property if a window is mapped without one.

The input field is used to communicate to the window manager the input focus model used by the client (see Section 3.4.1.7).

Clients with the Globally Active and No Input models should set the input flag to False. Clients with the Passive and Locally Active models should set the input flag to True.

From the client's point of view, the window manager will regard the client's top-level window as being in one of three states:

• Normal
• Iconic
• Withdrawn

The semantics of these states are described in 3.4.1.4. Newly created windows start in the Withdrawn state. Transitions between states happen when a top-level window is mapped and unmapped and when the window manager receives certain messages.

The value of the initial_state field determines the state the client wishes to be in at the time the top-level window is mapped from the Withdrawn state, as shown in the following table:

The icon_pixmap field may specify a pixmap to be used as an icon. This pixmap should be:

• One of the sizes specified in the WM_ICON_SIZE property on the root if it exists (see Section 3.4.1.3.2).
• 1-bit deep. The window manager will select, through the defaults database, suitable background (for the 0 bits) and foreground (for the 1 bits) colors. These defaults can, of course, specify different colors for the icons of different clients.

The icon_mask specifies which pixels of the icon_pixmap should be used as the icon, allowing for icons to appear nonrectangular.

The icon_window field is the ID of a window the client wants used as its icon. Most, but not all, window managers will support icon windows. Those that do not are likely to have a user interface in which small windows that behave like icons are completely inappropriate. Clients should not attempt to remedy the omission by working around it.

Clients that need more capabilities from the icons than a simple two-color bitmap should use icon windows. Rules for clients that do are set out in Section 3.4.1.9.

The (icon_x,icon_y) coordinate is a hint to the window manager as to where it should position the icon. The policies of the window manager control the positioning of icons, so clients should not depend on attention being paid to this hint.

The window_group field lets the client specify that this window belongs to a group of windows. An example is a single client manipulating multiple children of the root window.

Conventions

1. The window_group field should be set to the ID of the group leader. The window group leader may be a window that exists only for that purpose; a placeholder group leader of this kind would never be mapped either by the client or by the window manager.
2. The properties of the window group leader are those for the group as a whole (for example, the icon to be shown when the entire group is iconified).

Window managers may provide facilities for manipulating the group as a whole. Clients, at present, have no way to operate on the group as a whole.

The messages bit, if set in the flags field, indicates that the client is using an obsolete window manager communication protocols, ¹¹⁾ rather than the WM_PROTOCOLS mechanism of Section 3.4.1.2.7.

The UrgencyHint flag, if set in the flags field, indicates that the client deems the window contents to be urgent, requiring the timely response of the user. The window manager must make some effort to draw the user's attention to this window while this flag is set. The window manager must also monitor the state this flag for the entire time the window is in the Normal or Iconic state and must take appropriate action when the state of the flag changes. The flag is otherwise independent of the window's state; in particular, the window manager is not required to deiconify the window if the client sets the flag on an Iconic window. Clients must provide some means by which the user can cause the UrgencyHint flag to be set to zero or the window to be withdrawn. The user's action can either mitigate the actual condition that made the window urgent, or it can merely shut of the alarm.

Rationale

This mechanism is useful for alarm dialog boxes or reminder windows, in cases where mapping the window is not enough (e.g. in the presence of multi-workspace or virtual desktop window managers), and where using an override-redirect window is too intrusive. For example, the window manager may attract attention to an urgent window by adding an indicator to its title bar or its icon. Window managers may also take additional action for a window that is newly urgent, such as by flashing its icon (if the window is iconic) or by raising it to the top of the stack.

11) This obsolete protocol was described in the July 27. 1988 draft of the ICCCM. Windows using it can also be detected because their WM HINTS properties are four bytes Ionger than expected. Window managers are free to support clients using the obsolete protocol in a "backwards compatibility"mode._

3.4.1.2.5 WM CLASS Property

The WM_CLASS property (of type STRING without control characters) contains two consecutive null-terminated strings. These specify the Instance and Class names to be used by both the client and the window manager for looking up resources for the application or as identifying information. This property must be present when the window leaves the Withdrawn state and may be changed only while the window is in the Withdrawn state. Window managers may examine the property only when they start up and when the window leaves the Withdrawn state, but there should be no need for a client to change its state dynamically.

The two strings, respectively, are:

• A string that names the particular instance of the application to which the client that owns this window belongs. Resources that are specified by instance name override any resources that are specified by class name. Instance names can be specified by the user in an operating-system specific manner. On POSIX-conformant systems, the following conventions are used:

  -	If "-name NAME" is given on the command line, NAME is used as the instance name.
  -	Otherwise, if the environment variable RESOURCE_NAME is set, its value will be used as the instance name.
  -	Otherwise, the trailing part of the name used to invoke the program (argv[0] stripped of any directory names) is used as the instance name.

• A string that names the general class of applications to which the client that owns this window belongs. Resources that are specified by class apply to all applications that have the same class name. Class names are specified by the application writer. Examples of commonly used class names include: "Emacs", "XTerm", "XClock", "XLoad", and so on.

Note that WM_CLASS strings are null-terminated and, thus, differ from the general conventions that STRING properties are null-separated. This inconsistency is necessary for backwards compatibility.

3.4.1.2.6 WM TRANSIENT FOR Property

The WM_TRANSIENT_FOR property (of type WINDOW) contains the ID of another top-level window. The implication is that this window is a pop-up on behalf of the named window, and window managers may decide not to decorate transient windows or may treat them differently in other ways. In particular, window managers should present newly mapped WM_TRANSIENT_FOR windows without requiring any user interaction, even if mapping toplevel windows normally does require interaction. Dialogue boxes, for example, are an example of windows that should have WM_TRANSIENT_FOR set.

It is important not to confuse WM_TRANSIENT_FOR with override-redirect. WM_TRANSIENT_FOR should be used in those cases where the pointer is not grabbed while the window is mapped (in other words, if other windows are allowed to be active while the transient is up). If other windows must be prevented from processing input (for example, when implementing pop-up menus), use override-redirect and grab the pointer while the window is mapped.

3.4.1.2.7 WM PROTOCOLS Property

The WM_PROTOCOLS property (of type ATOM) is a list of atoms. Each atom identifies a communication protocol between the client and the window manager in which the client is willing to participate. Atoms can identify both standard protocols and private protocols specific to individual window managers.

All the protocols in which a client can volunteer to take part involve the window manager sending the client a ClientMessage event and the client taking appropriate action. For details of the contents of the event, see Section 3.4.2.8. In each case, the protocol transactions are initiated by the window manager.

The WM_PROTOCOLS property is not required. If it is not present, the client does not want to participate in any window manager protocols.

The X Consortium will maintain a registry of protocols to avoid collisions in the name space. The following table lists the protocols that have been defined to date.

It is expected that this table will grow over time.

3.4.1.2.8 WM COLORMAP WINDOWS Property

The WM_COLORMAP_WINDOWS property (of type WINDOW) on a top-level window is a list of the IDs of windows that may need colormaps installed that differ from the colormap of the top-level window. The window manager will watch this list of windows for changes in their colormap attributes. The top-level window is always (implicitly or explicitly) on the watch list. For the details of this mechanism, see Section 3.4.1.8.

3.4.1.2.9 WM CLIENT_MACHIE Property

The client should set the WM_CLIENT_MACHINE property (of one of the TEXT types) to a string that forms the name of the machine running the client as seen from the machine running the server.

3.4.1.3 Window Manager Properties

The properties that were described in the previous section are those that the client is responsible for maintaining on its top-level windows. This section describes the properties that the window manager places on client's top-level windows and on the root.

3.4.1.3.1 WM STATE Property

The window manger will place a WM-STATE property (of type WM-STATE) on each top-level client window that is not in the Withdrawn state. Top-level windows in the Withdrawn state may or may not have the WM_STATE property. Once the top-level window has been with-drawn, the client may re-use it for another purpose. Clients that do so should remove the WM_STATE property if it is still present.

Some clients (such as xprop) will ask the user to click over a window on which the program is to operate, Typically, the intent is for this to be a top-level window. To find a top-level window, clients should search the window hierarchy beneath the selected location for a window with the WM_STATE property. This search must be recursive in order to cover all window manager reparenting possibilities. If no window with a WM_STATE property is found, it is recommended that programs use a mapped child-of-root window if one is present beneath the selected location.

The contents of the WM_STATE property are defined as follows:

The following table lists the WM_STATE state values:

Adding other fields to this property is reserved to the X Consortium. Values for the state field other than those defined in the above table are reserved for use by X Consortium.

The state field describes the window manger's idea of the state the window is in, which may not match the client's idea as expressed in the initial_state field of the WM_HINTS property (for example, if the user has asked the window manager to iconify the window). If it is NormalState, the window manager believes the client should be animating its window. If it is IconicState, the client should animate its icon window. In either state, clients should be prepared to handle exposure events from either window.

When the window is withdrawn, the window manager will either change the state field's value to WithdrawnState or it will remove the WM_STATE property entirely.

The icon field should contain the window ID of the window that the window manager uses as the icon for the window on which this property is set. If no such window exists, the icon field should be None. Note that this window could be but is not necessarily the same window as the icon window that the client may have specified in its WM_HINTS property. The WM_STATE icon may be a window that the window manager has supplied and that contains the client's icon pix-map, or it may be an ancestor of the client's icon window.