Network Working Group Lorenzo Aguilar Request for Comments: 965 SRI International December 1985 A Format for a Graphical Communication Protocol STATUS OF THIS MEMO This paper describes the requirements for a graphical format on which to base a graphical on-line communication protocol. The proposal is an Interactive Graphical Communication Format using the GKSM session metafile. Distribution of this memo is unlimited. ABSTRACT This paper describes the requirements for a graphical format on which to base a graphical on-line communication protocol. It is argued that on-line graphical communication is similar to graphical session capture, and thus we propose an Interactive Graphical Communication Format using the GKSM session metafile. We discuss the items that we believe complement the GKSM metafile as a format for on-line interactive exchanges. One key application area of such a format is multi-media on-line conferencing; therefore, we present a conferencing software architecture for processing the proposed format. We make this format specification available to those planning multi-media conferencing systems as a contribution toward the development of a graphical communication protocol that will permit the interoperation of these systems. We hope this contribution will encourage the discussion of multimedia data exchange and the proposal of solutions. At SRI, we stay open to the exploration of alternatives and we will continue our research and development work in this problem area. ACKNOWLEDGEMENTS The author wants to thank Andy Poggio of SRI who made many insightful and valuable suggestions that trimmed and improved level U. His expertise in multi-media communication systems and his encouragement were a most positive input to the creation of this IGCF. Dave Worthington of SRI also participated in the project discussions involving this IGCF. Thanks are also due to Tom Powers, chairman of ANSI X3H33, who opened this forum to the presentation of an earlier version of this paper, thereby providing an opportunity for the invaluable feedback of the X3H33 members. Jon Postel of ISI recommended a number of changes that made this paper more coherent and accessible. Aguilar [Page 1] RFC 965 December 1985 A Format for a Graphical Communication Protocol Most of the work reported in this paper was sponsored by the U.S. Navy, Naval Electronic Systems Command, Washington D.C., under Contract No. N00039-83-K-0623. I. INTRODUCTION A. Use of a Graphical Communication Protocol In the field of computer communications, a protocol is a procedure executed by two cooperating processes in order to attain a meaningful exchange of information. A graphical communication protocol is needed to exchange interactive vector graphics information, possibly in conjunction with other information media like voice, text, and video. Within this multi-media communication environment, computer vector graphics plays a key role because it takes full advantage of the processing capabilities of communicating computers and human users, and thus it is far more compact than digital images which are not generated from data structures containing positional information. Vector graphical communication trades intensive use of storage and processing, at the communicating ends, in return for a low volume of exchanged data, because workstations with graphical hardware exchange graphics commands in conjunction with large data structures at the transmitter and receivers. In this manner, the transmission of a single command can produce extensive changes in the data displayed at the sending and receiving ends. It is helpful to situate the aforesaid protocol at one of the functional levels of the ISO Open Systems Interconnection Reference Model [1]. Within such a model, a graphical protocol functionality belongs primarily in the application level, though some of it fits in the presentation level. We can distinguish the following components of a communication protocol: a) a data format b) rules to interpret transmitted data c) state information tables d) message exchange rules A format for a graphical protocol should provide the layout of the transmitted data, and indicate how the formated data are associated with interpretation rules. The choice of format influences the state tables to be maintained for the correct processing of the transmitted data stream. The graphical format has a minor influence on the exchange rules, which should provide for the efficient use of transmission capacity to transport the data under such a format. Besides the graphical format, there are Aguilar [Page 2] RFC 965 December 1985 A Format for a Graphical Communication Protocol other aspects of a graphical protocol that determine state tables and exchange rules. This paper concentrates in the data format, and it does not discuss the message exchange. Nevertheless, we discuss a simple software architecture for generating and interpreting data streams written in our proposed format. Further, we give an example of an application of a proposed format (in Appendix B), and it illustrates the type of message exchanges that are needed for establishing a communication session and exchanging graphical information. Those in the computer communication field are well aware of the importance of widely accepted protocols in order to achieve meaningful communication. Those who need to implement interactive graphical communications today are confronted with the lack of an standard for computer graphics communication among application programs. Nevertheless, we can use some of the work already done by the computer graphics standard bodies. As a matter of fact, ISO and ANSI have already appended, to the Graphical Kernel System (GKS) standard, the GKSM session metafile specification that has many of the features needed for an on-line graphical protocol. It is pertinent to mention an example of graphical communication that illustrates the real-time nature of the interaction and also illustrates the use of graphics in conjunction with other information media. With audio-graphics conferencing, a group of individuals at two or more locations can carry on an electronic meeting. They can converse over voice channels and concurrently share a graphics space on which they can display, point at, and manipulate vector graphics pictures [2, 3, 4, 5, 6, 7]. The conference voice channels can be provided by a variety of transmission technologies. The shared graphics space can be implemented on workstations that display the pictures and permit graphical interaction and communication with other locations. The communication of operations upon pictures involves modifications to the underlying data structures, but we are concerned with graphical database updating only to the extent that such updating supports the communication. In order to play out a recorded graphical session, we will need indications of the rate at which the graphical elements must be shown and the graphical operations recreated. We do not include the means for indicating the timing of a session in a format because our main purpose is to use it in mixed-media communication environments. In these environments, the play-out timing must be compatible across information media in order to coordinate them. Therefore, we leave the timing mechanisms to conference-control Aguilar [Page 3] RFC 965 December 1985 A Format for a Graphical Communication Protocol modules. We also leave to conference control processes the manner in which a conferee station emulates a graphical capability that it lacks. One example is the representation of color in monochrome displays. B. Relationship to Other Work There are a number of actual, and proposed, standards for graphics information exchange. In the following, we explain the reasons why, at present, none of them can be used as the basis of an on-line protocol. As some of these standards evolve, however, some may become suitable. Moreover, the experience gained with early on-line graphics communication systems will provide insight into the proper standard extensions to support more advanced systems. Such insight could also be used to modify the format proposed in this paper, which we consider an initial approach to the problem. In the future, the format proposed in this paper could be replaced by one of the aforesaid extended standards. The North American Presentation Level Protocol Syntax, NAPLPS, specifies a data syntax and application semantics for one-way teletext information dissemination and two-way videotex database access and transaction services. The two-way videotex operational model is based on the concept of a consumer and an information provider or service operator. Because of this asymmetry, it is assumed that almost all graphical information will flow from the provider toward the consumer. In the reverse direction, the consumer is expected to manipulate and transmit alphanumeric information, for the most part. Although this standard includes geometric drawing primitives, a user cannot directly modify shapes drawn with the primitives. At present, NAPLPS does not include interaction concepts like picture transformations or detectability, which are fundamental for attaining a shared graphical workspace. Neither does it allow key graphics input devices like mice, joysticks, stylus, rotating balls, or light pens, which are needed for simple and efficient editing of the shared workspace. We want to have user-to-user graphical communication that features the level of sophistication and ease of interaction provided by today's interactive graphics packages. Computer vector graphics can provide both because its paradigm includes an application program that keeps track of a very large number of possible changes of state of the displayed picture. In addition, the application drives a powerful graphics package, like GKS or ACM Core. In the videotex paradigm, the provider application only Aguilar [Page 4] RFC 965 December 1985 A Format for a Graphical Communication Protocol allows limited changes to the displayed image, primarily database retrieval requests. Also, the paradigm does not include a separate graphics package. Both the graphics functionality and the data format are collapsed into a coding specification, like NAPLPS. In this paper we are interested primarily in business and industrial applications where there is a two-way, or multi-way, flow of vector graphics information among the users. The users will have workstations with substantial processing and storage capacities, and high-resolution monitors; moreover, the communication will be on a distributed architecture not depending on a central server host, like the provider application host of videotex. Currently, the videotex equipment at the consumer end consists of inexpensive microprocessor-based decoders or personal computer boards driving, in most cases, low-resolution standard TV sets and personal computer displays. There is already affordable technology to produce sophisticated decoders and high-resolution graphics devices. The videotex standards need extensive revisions to take advantage of these advances; in particular, they should consider the receiving devices as capable of hosting a programmable customer-application process. When this happens, videotex protocols will be applicable to our intended problem areas [8]. The Computer Graphics Metafile [9] will become an international and North American standard for graphics picture interchange in the near future. However, the CGM, also referred as VDM, is a picture-capture metafile that only records the final result of a graphics session. It is not intended to record the picture-creation process, which is fundamental for the interactive applications that we are addressing. Moreover, the CGM is presently aimed at a minimum support of GKS functionality. It will be some time before the CGM will have some of the elements needed for on-line interaction. If, after these additions, the CGM is augmented for session capture, it would become a logical candidate for a protocol format. Another future standard is the Computer Graphics Interface, CGI also referred as VDI [10]. The CGI is a standard functional and syntactical specification of the control and data exchange between device-independent graphics software and one or more device-dependent graphics device drivers. A major use of the CGI is for the communication between an application host and a graphics device, but the asymmetry between its intended communicating ends hinders the use of CGI for our purposes. Aguilar [Page 5] RFC 965 December 1985 A Format for a Graphical Communication Protocol As previously stated, we want to take advantage of intelligence and storage at the communicating ends in order to achieve powerful information-conveying effects using narrow-bandwidth channels. This requires that the format we seek must have items for communication between two applications. In contrast, the CGI streams are processed by device-dependent drivers, rather than by applications. The CGI specification does include application data elements, but only to be stored in a metafile. These application data elements are not interpreted by the drivers, but by applications that read the metafile, some time after metafile creation. Furthermore, the CGI has elements for obtaining graphical input, as well as elements for inquiring graphics device capabilities, characteristics, and states. Later, in Section III, we explain why these two classes of elements are unnecessary for the communication protocol we need. As the CGI evolves, it will undergo significant changes, and, in the future, it may become a very suitable kernel for the graphics protocol we seek. As a matter of fact, the CGI will be the communication protocol between graphical application hosts and graphics terminals. At SRI we are tracking its evolution, and we are interested in defining a format based on the CGI. Finally, the Initial Graphics Exchange Specification [11] is not aimed at our primary area of interest. The IGES defines standard file and language formats for storing and transmitting product-definition data that can be used, in part, to generate engineering drawings and other graphical representations of engineering products. Besides the CAD orientation of IGES, the graphical output function may be secondary to other goals like transmitting numerical-control machine instructions. II. OPERATIONAL REQUIREMENTS AND USABILITY The main goal of this paper is to lay the groundwork for the development of a vector graphics format to be used as a basis for an on-line graphical communication protocol. We call such a format an "interactive graphical communication format," or IGCF. In this section we describe some operational requirements and usable characteristics for an IGCF. A. Interoperation of Heterogeneous Systems A first functional requirement is that an IGCF must permit communication among heterogeneous graphical systems differing both in the hardware used and in the software of their graphics Aguilar [Page 6] RFC 965 December 1985 A Format for a Graphical Communication Protocol application interfaces. This is a fundamental for attaining communication among similar graphical application programs running on dissimilar hardware and using dissimilar graphics interface packages. Some examples of such application programs are graphics editors, CAD systems, and graphical database retrieval programs communicating with other editors, CAD programs, and graphical databases, respectively. B. Picture Capture A required characteristic of an IGCF is that it must be usable for the exchange of static graphic pictures, i.e. for picture capture; yet, it must not be restricted to final picture recording only. There will be picture exchanges as part of the interactive communication, and we anticipate the need to record the state of a picture at some points during the on-line graphics engagement. We foresee the creation of graphical IGCF libraries containing object definitions and pictures for inclusion in new pictures. Since metafiles have been used for a long time to capture pictures, there is a strong motivation to base an IGCF on a metafile standard in order to secure compatibility with a large number of metafile sources and consumers. C. Prompt Transmission In some forms of interactive graphical communication, like audiographics conferencing, it is critical to convey across users the real-time nature of the interaction. This dictates that object creations and manipulations be transmitted as they happen rather than as a final result since a substantial part of the information may be transmitted concurrently with the construction or operation of an object, possibly through associated media like voice. Since both construction and manipulation processes have to be transmitted, there is a limit to the number of intermediate states that can be economically transmitted. A third requirement is, therefore, that the IGCF elements provide fine "granularity" to convey the dynamics of the constructions and manipulations. We believe that it is sufficient that the IGCF have basic construction elements like polygons, markers, polylines, and text strings and that it transmit them only when they are completed; i.e., it is not necessary to transmit partial constructions of such elements. The problem for manipulations extends beyond an IGCF. Whereas we know that an IGCF should include segment transformations, segment highlighting and segment visibility on/off, the transmitter must Aguilar [Page 7] RFC 965 December 1985 A Format for a Graphical Communication Protocol decide how often to sample an on-going transformation and transmit its current state. The choice of a sampling frequency will depend on the available transmission bandwidth. D. Low Traffic Volume In many of the applications we envision, coordinate graphics will be transmitted over narrow bandwidth channels, and thus it is essential to minimize traffic. Accordingly, several requirements are imposed on an IGCF to take advantage of the characteristics of the graphics communication intercourse and architecture in order to minimize traffic. An IGCF can help reduce traffic by including the basic geometric objects from which so many other objects are built. Moreover, an IGCF should permit the use of objects for the creation of more complex objects; since reuse is very common, the result is a reduction of traffic and storage cost. E. Preservation of Application Semantic Units A related requirement is that an IGCF must include elements to represent graphical objects corresponding to real world entities of the intended applications. For example, in a Navy application, the entities of interest are carriers, submarines, planes, and the like. We want to communicate such semantic units across systems and to treat them as unitary objects because, in many applications, communication is based on creating and operating such units. If an IGCF has elements to represent such semantic units, the communication traffic decreases because the entity definitions can be transmitted only once and then reused, and because the entities are manipulated as units rather than separately manipulating their components. It turns out that there is a small set of primary operations that can be applied to a graphical object, and an IGCF must have elements representing such operations. In contrast to dumb graphics terminals receiving screen refresh information from a host, we foresee graphical communication taking place among intelligent workstations that can exchange encoded operations, interpret them, and apply them to objects stored locally. F. Transmission Batching We previously indicated the desirability of conveying to the human users the real-time tempo of interactive graphics exchanges. However, it is possible to do so without having to transmit Aguilar [Page 8] RFC 965 December 1985 A Format for a Graphical Communication Protocol immediately all IGCF elements. As a matter of fact, IGCF elements should be divided into those causing a change on a displayed picture and those that do not, although both classes may cause changes to the stored graphical data structures. It is only necessary to transmit immediately those elements causing a visible change on a displayed picture because they are the ones whose reception and interpretation delivers information to a human user. The second class of elements can be batched and queued for transmission until one element of the first class is submitted. We call the first class update Group-1, and the second, update Group-2. The aforesaid division is quite important for packet communications because each packet contains a hefty amount of overhead control traffic. It is therefore mandatory to batch, into a packet, as much client data as possible in order to reduce total traffic. The batching units can be varied in size according to the network traffic and response time of conference hosts. During congested periods, the units may have to be increased, thus lowering the number of messages, and then reduced when congestion eases, thus increasing the number of messages. G. Simple Translation Between IGCF and User Interface According to the first requirement, an IGCF must permit the interoperation of related heterogeneous graphics applications. Such interoperation has, as an objective, the communication between human users or between a human and a database. Correspondingly, the interoperation involves a mapping between the user interface commands and the IGCF elements. It is not advisable to use the commands themselves as the IGCF elements; otherwise the exchange would depend on the communicating systems, and every pair of communicating systems would require an ad-hoc protocol. An additional usability characteristic is that there must be a simple mapping between IGCF elements and the actions represented by the user interface commands employed for graphical communications. This simplicity is a must because every communicating graphical system must have a translator that ideally should be very simple. It seems that the inclusion of command sequence delimiters in the IGCF helps the simplicity since the delimiters permit keeping a smaller amount of state information for processing an IGCF stream. We have verified the mapping from one set of commands for audiographics conferencing to the IGCF proposed in this paper. The Aguilar [Page 9] RFC 965 December 1985 A Format for a Graphical Communication Protocol mapping from user interface commands to IGCF can be done in a direct and efficient manner; on the other hand, the reverse mapping, from IGCF to user interface commands, is a more difficult task. We anticipate that, in order to improve performance, we will have to map the IGCF elements to calls to lower level subroutines implementing the user interface actions. Whereas such mapping is conceptually no more complex than translating IGCF to the commands themselves, it will require considerably more programming. III. ELEMENTS OF AN IGCF IGCF Element Classes In this section we list the classes of elements that we believe an IGCF should have in order to exchange vector graphics under the requirements of the previous section. The classes correspond to the common function classes in computer graphics interfaces, and each contains elements corresponding to interface primitives and attributes. We do not list the elements for each class because they are exemplified by the elements in the proposed IGCF. In the following list, two categories of functions are missing: functions used to query the status of a graphics system, and input functions. As a matter of fact, an IGCF only needs to have elements representing actions that cause a change in the state of the communicating graphical systems, and the inquire functions obviously do not change their state. Even though an input function executed at the transmitting end causes a local change, it is not necessary to transmit the input command itself. The receivers only need to get the data input, in IGCF representation, and they can process the data in any manner, maybe simulating local input actions. Control Elements for workstation: initialization, control and transformation; and elements for normalization transformation. (The normalization and workstation transformations can be used to implement zooming.) Primitive attributes Elements for primitive, segment, and workstation attributes. Output primitives Elements for output primitives. Aguilar [Page 10] RFC 965 December 1985 A Format for a Graphical Communication Protocol Segmentation Elements for basic segmentation and workstation independent segment storage. Object manipulations can be implemented with segment transformations. Object insertion can be implemented using segment recall and segment visibility. Object deletion can be implemented using segment deletion and segment visibility. Object selection can use segment highlighting as feedback to the user. Dynamics A considerable part of the graphical information exchanged through an IGCF will be in the form of pointer movements over a background picture. Pointer tracking is used to transmit points sampled from a graphical pointer trace in order to reproduce, at the receivers, the movement of the pointer at the sender site. This can be done either by just moving the cursor or by tracing its movement with a line. Rubber band echoes are used to signal areas, routes, and scopes in a highly dynamic way. These are indicated by an echo reference point and a feedback point. Hierarchical object definitions The requirement for preserving application semantics dictated that an IGCF include the means to represent objects that stand for application entities, and to manipulate such entities as graphical units. Furthermore, the low-traffic-volume requirement called for the use of already existing objects for the creation of new ones. One way to meet the aforesaid requirements is by including in an IGCF the means to represent object hierarchies. In such a hierarchy an object is a set of output primitives associated with a set of attribute values or a set of lower-level objects, each associated with a composition of transformations [12]. Graphics segments can be used to implement objects in the lowest level of a hierarchy. The definition of a higher-level object can be represented by sequences of IGCF elements describing the definition process. Such a definition can be done by instantiating lower-level objects with specific transformation parameters. Thus an IGCF must incorporate brackets to mark the beginning and end of object definitions, object instantiations, and object redefinitions. Aguilar [Page 11] RFC 965 December 1985 A Format for a Graphical Communication Protocol In order to complement the mechanism for object definition, an IGCF must permit the use of a flexible alphabet for creating object identifiers that ensure the uniqueness of an identifier in a hierarchy. The construction of the object identifiers is not part of an IGCF, an IGCF only has to represent the identifiers. Further, an identifier has to be independent of a communication session and a particular graphics system so that identifiers created at a host during one session can be used, in other sessions possibly involving other hosts, to recall the objects they label. We also leave to the communicating systems the implementation of mechanisms to resolve duplicate identifiers when merging two hierarchies, created in different sessions. In this paper we shall limit ourselves to the warning that segment numbers do not qualify as identifiers because they depend on the session and state of the system in which they are created. In addition to object definition and instantiation, an IGCF should have elements representing operations on objects. The operations so far identified are: transformation, deletion, display, disappearance, expose, and hide. Expose is used to uncover objects on a screen that are hidden by other objects; hide is used to place an object behind others on a screen. IV. A PROPOSED IGCF A. Using the GKSM as a Basis An IGCF must be usable to transmit all graphical actions in a conference session. This suggests to base an IGCF on a standard session-capture graphics metafile, thus ensuring compatibility with a large user population. We have based the proposed IGCF, PIGCF, on the GKSM session-capture metafile specification because GKSM contains many of the elements identified for an IGCF [14]. In addition, the audit trail orientation of GKSM permits the recording of interactive communication sessions for later play out, and this is a feature that we anticipate will be frequently used. The GKSM is a proper subset of our PIGCF and thus any graphical system developed to handle the PIGCF, can read a GKSM metafile. Conversely, the applications using the PIGCF should have an option for constraining session recording only to the GKSM part, possibly suppressing some session events. By doing so, we will be able to ship a GKSM metafile to any correspondent who has GKSM Aguilar [Page 12] RFC 965 December 1985 A Format for a Graphical Communication Protocol interpretation software. Alternatively, an application with a GKSM interpreter but without an PIGCF interpreter can read a PIGCF file interpreting only the GKSM part and ignoring the rest. Whereas the GKSM was specified for the GKS system, we believe that the GKSM is a sound and general basis for all of our 2-D applications. We feel that the GKSM specification is not parochial to GKS systems but contains all the most useful items desired in a metafile. In the future, we expect to tackle applications requiring 3-D, like interactive repair and maintenance aids. When GKS be augmented with 3-D capabilities [13], we will extend the PIGCF with any necessary elements. We are aware that the GKSM specification is not part of the GKS standard itself but is an appendix recommending such a metafile format. Nevertheless, all the GKS vendor implementations that we know of, at the present time, support GKSM metafile output and interpretation. If this trend continues, as we expect, we will be able to exchange graphical files with a large base of GKS installations. There will indeed be many of them since GKS will be adopted as an standard by ISO and by many national standard bodies in the near future. B. Positional Information Coordinates Following the GKSM convention, the PIGCF positional information is in normalized device coordinates, NDC. Thus the originator of a conference must indicate the workstation window for the conference. This window is the sub-rectangle of the NDC space enclosing the area of interest for the conference. In most cases, the participating workstations will take this window as their own. However, the graphical systems should provide for the possibility of a workstation choosing a different workstation window, which may contain the conference window or just overlap it. Except for special cases, a conference originator should not state a conference workstation viewport. In this manner, each workstation can display its workstation viewport in the most convenient portion of the screen. There will be conferences where the participating workstations will maintain the positional information in world coordinates, WC. It might be necessary to reconstruct the world dimensions after transmission because such dimensions have a relevant meaning for the application, like sizes of components or distances. In this case, a workstation will have to map from WC to NDC before transmitting and from NDC to WC after receiving. At the outset, the conference originator has to specify the world window and the Aguilar [Page 13] RFC 965 December 1985 A Format for a Graphical Communication Protocol NDC viewport used in the conference in order for the conferencing workstations to do such mappings. These mappings could be done by the presentation layer, in terms of the ISO Open Systems Interconnection Reference Model, in a manner that is transparent to the communicating application programs. Most often all workstations will have the same world windows and NDC viewports. However, the graphical systems will provide for the possibility of a workstation choosing a different window or viewport, but such workstation will have to record the conference ones for doing the aforesaid mappings. There are graphical systems, like the ACM Core, that do not provide for a workstation transformation. In such systems, the NDC viewport is considered to be the workstation window for the aforesaid mappings. C. Layers of the PIGCF There are two levels in the PIGCF a lower level L and an upper one U. The lower level L is just the GKSM metafile specification as defined in Appendix E of the proposed GKS ANSI standard [14]. We have excerpted most of Appendix E of [14] at the end of this RFC as our Appendix A. All level L elements belong to the update Group-1 except: SET DEFERRAL STATE, the output primitive attribute elements, the workstation attribute elements, CLIPPING RECTANGLE, CREATE SEGMENT, CLOSE SEGMENT, RENAME SEGMENT, SET SEGMENT PRIORITY, and SET DETECTABILITY. The upper level U is those elements that we believe complement the GKSM for general on-line graphical exchanges. This layering conforms to the graphics metafile level-structure described in Enderle et. al [15]. Under such structuring, an application oriented metafile can be based on graphical metafiles. D. PIGCF Elements in the Level U The level U items are encoded as GKSM user item elements so that a PIGCF file will conform to the GKSM metafile specification. Accordingly, a PIGCF file will be a GKSM metafile in its entirety. We use the same formatting conventions as the GKSM specification. Those unfamiliar with these conventions should read the beginning of the appendix. The following items belong to the second update group: the two items for object definition, the two items for object redefinition, the two items for object instantiation, the two items for normalization transformation, SELECT COMPONENT, and RECALL LIBRARY. The remaining items belong to the first update group. Aguilar [Page 14] RFC 965 December 1985 A Format for a Graphical Communication Protocol Items for Object Definition BEGIN DEFINITION | 'GKSM 120' | L | Indicates beginning of object definition sequence END DEFINITION | 'GKSM 121' | L | I | Indicates end of object definition sequence. I(Nc): object identifier ( N preceding c, i, r means an arbitrary number of characters, integers, or reals.) Objects defined interactively are made visible on the screen; i.e. they are automatically instantiated. If only the definition is to be kept but not the image, a DISAPPEAR item must follow. BEGIN REDEFINITION | 'GKSM 122' | L | I | Indicates beginning of object redefinition sequence I(Nc): object identifier END REDEFINITION | 'GKSM 123' | L | Indicates end of object redefinition sequence Items for Object Instantiation BEGIN INSTANTIATION | 'GKSM 124' | L | I | Indicates beginning of object instantiation sequence I(Nc): Object identifier END INSTANTIATION | 'GKSM 125' | L | Indicates end of object instantiation sequence Aguilar [Page 15] RFC 965 December 1985 A Format for a Graphical Communication Protocol Items for Object Manipulation TRANSFORM OBJECT | 'GKSM 126' | L | C | I | M | Apply transformation M to object I C: number of characters in identifier I(Nc): object id M(6r): upper and center rows of a 3x3 matrix representing a 2D homogeneous transformation [12]. M 11 M 12 M 13 M 21 M 22 M 23 DELETE OBJECT | 'GKSM 127' | L | I | I(Nc): object identifier DISPLAY OBJECT | 'GKSM 128' | L | I | Turn on visibility of object I I(Nc): object identifier DISAPPEAR OBJECT | 'GKSM 129' | L | I | Turn off visibility of object I I(Nc): object identifier EXPOSE OBJECT | 'GKSM 130' | L | I | Redisplay object I on top of any overlapping objects I(c): object identifier HIDE OBJECT | 'GKSM 131' | L | I | Redisplay object I behind any overlapping objects I(c): object identifier Aguilar [Page 16] RFC 965 December 1985 A Format for a Graphical Communication Protocol SELECT COMPONENT | 'GKSM 132' | L | I | P | Select component P of object I I(c): object identifier P(i): pick id of component This is used to select a group of output primitives identified by P in a segment associated with I. ERASE COMPONENT | 'GKSM 133' | L | I | P | Erase component P of object I I(c): object identifier P(i): pick id of component This erases a group of output primitives identified by P in a segment associated with I. This element can be used only within a REDEFINE OBJECT sequence. Items for Normalization Transformation SET WINDOW | 'GKSM 134' | L | W | Define boundaries of world window for normalization transformation. W(4r): limits of world window (XMIN, XMAX, YMIN, YMAX ) SET VIEWPORT | 'GKSM 135' | L | V | Define boundaries of NDC viewport for normalization transformation. V(4r): limits of NDC viewport (XMIN, XMAX, YMIN, YMAX ) Aguilar [Page 17] RFC 965 December 1985 A Format for a Graphical Communication Protocol Items for Other Operations ABORT | 'GKSM 136' | L | Abort ongoing operation transmitted in PIGCF stream. This provides the means to abort unwanted or erroneous operations. Only the innermost operation of a nested sequence is aborted; successive aborts can be used to get out of several levels of operation nesting. POINTER TRACKING | 'GKSM 137' | L | T | P | Update graphical pointer position to P T(i): 0 causes only cursor to be moved 1 causes cursor movement to be traced with a line P(p): a point sampled from graphical pointer movement trace Aguilar [Page 18] RFC 965 December 1985 A Format for a Graphical Communication Protocol RUBBER BAND | 'GKSM 138' | L | T | P | Echo a rubber band of type T with given reference and feedback points. The first occurrence of this item in a sequence carries the coordinates of the echo reference point. Subsequent occurrences carry updates to a pointer position indicating an echo feedback point. T(i): echo type ( 0 echo reference point; > 0 echo feedback: 1 = line, 2 = rectangle, 3 = circle ) P(r): echo reference point (T = 0), or echo feedback point (T > 0) The reference and feedback points are: T = 1 - reference is one end of line, feedback is other end. T = 2 - reference is one corner of rectangle, feedback is opposite corner. T = 3 - reference is center of circle, feedback is perimeter point. RECALL LIBRARY | 'GKSM 139' | L | F | Recall graphical library in file F F(i): name of file containing library The graphical pictures in F and all their components become available for use during the communication session. The pictures are assumed to be recorded with the PIGCF, and their components have to be displayed with DISPLAY OBJECT elements or similar actions so that the pictures become visible. Aguilar [Page 19] RFC 965 December 1985 A Format for a Graphical Communication Protocol V. AN ARCHITECTURE FOR PIGCF PROCESSING This section presents an example software architecture for the generation and interpretation of PIGCF in a multimedia conferencing system using GKS as the underlying programmer's graphics interface. This section should not be interpreted as a definitive statement of such an architecture, but only as an exercise to illustrate how the format proposed in this paper fits within the overall framework of a conferencing system. Choosing GKS simplifies the example architecture; nevertheless, other graphics packages can be used by adding, to the architecture, the modules to interpret and generate the PIGCF level L items. Figure 1 shows the major software modules charged with graphics interaction and display at a conferencing workstation. This is a familiar programmer's view of the graphics pipeline. A conferencing application program updates data structures and uses device-independent graphics services through a language binding. These services, in turn, use device-dependent graphics services that call on device drivers to accept input and to present graphic pictures. The application performs numerous other functions for conference management and control of other media streams, but we need not consider them in this example. In Figure 2, the basic graphics pipeline has been augmented with the software modules involved in the generation, transmission, reception, and interpretation of PIGCF streams. The application has a module for interpreting the lower and higher levels of PIGCF and one for generating the upper level U. The device-independent graphics services include modules for generating and interpreting the lower level, L. This reflects the current practice of including the generation and interpretation functions in the graphics package. There is also a module that transmits the outgoing PIGCF streams to remote work stations. Similarly, there is a module that receives incoming streams from remote stations. In actual practice, the transmit and receive modules are decomposed into several processes implementing a layered protocol architecture. A process receives both levels of PIGCF and writes them into a conference record metafile for future use. A router process receives and forwards PIGCF traffic from and to the modules previously referred. This router is likely to be replaced by independent communication interfaces between pairs of modules exchanging PIGCF. The thick arrows show the flow of outgoing PIGCF, whereas the thin arrows show the incoming PIGCF flow. We first follow the outgoing path, starting at the application. The application processes local user actions which are transfo