zhangj@igd.fhg.de, http://www.igd.fhg.de/~zhangj/me.html
Martin Fruehauf, Computer Graphics Center (ZGDV), Dept. Mobile Information Visualization, Wilhelminenstr. 7, 64283 Darmstadt, Germany,
fruehauf@igd.fhg.de, http://www.igd.fhg.de/~fruehauf/me.html
Thomas Kirste, Computer Graphics Center (ZGDV), Dept. Mobile Information Visualization, Wilhelminenstr. 7, 64283 Darmstadt, Germany,
kirste@igd.fhg.de, http://www.igd.fhg.de/~kirste/me.html
Compressed postscript version of this paper for printing
A natural benefit of using URL is that it supports universal client. URL denotes not only a document address but also method to access it. A URL implementation is responsible to access a document, while the client on it must understand and visualize it. Based on a URL implementation, it is therefore not difficult to build a universal client. CERN has released an implementation of URL, the libwww.a. Besides providing URL based document access, libwww.a also translates representation of documents from different information services into HTML, so a universal client on it needs only to understand HTML. Xmosaic from the NCSA is an example of such a universal client.
In fact, general third-party applications are also capable of accessing the universe in the same way. This is important, because universal client has usually a restricted functionality and works monolithicly. Through Xmosaic, for example, a third-party application is able to access WWW document with a URL, but human intervention is required to save the document that Xmosaic fetches locally for further processing. When third-party applications do this themselves, the access will be seemless.
The significance of this is WWW may be used by different applications, distributed across the Internet as a world-wide, large-scale information base, on which they can carry out information sharing, information exchange and co-operative work [frivold] .
This figure illustrates: WWW constructs a sharable information universe which is accessible to universal clients and general third-party applications through libwww.a
HyperPicture provides the data model, architectural support and functionality to create and manipulate arbitrary hypermedia-structures. It simplifies the realization of complex interactive applications that require a broad range of different object and media types (such as raster images, graphics, text, sound and video), the administration of very large amounts of data, the facilities for the easy construction of complex data structures, and support of interactive navigation within these structures.
The architectural concept of HyperPicture defines the following central system components: Storage Management, Object Management, Session Management and Presentation and Interaction Modules (PIM). Here we describe PIMs in some detail, we use Xmosaic as an external PIM.
PIMs manage the type specific presentation of data objects, interpretation of event- and extent-definitions and the generation of events. In addition, they provide object-type specific mechanisms for object presentation and manipulation. external applications such as a CAD-package may be used by the system as PIMs, allowing HyperPicture to operate, in effect, as an application integrator .
The support of external editors and data processing software makes HyperPicture an open HyperPicture System.
For two hypertext systems that interoperate with each other, it is much more challenging to realize link construction and traversal across them than simply to realize data sharing or exchange between them [carr] . So in the preliminary scenario , we intend to support bidirectional link construction and traversal across HyperPicture and WWW.
This consists of two parts of work: supporting link from the HyperPicture to the WWW, so that the user is able to jump into Web from HyperPicture; and supporting link from the WWW to the HyperPicture, therefore the user is able to visit HyperPicture information structure from the WWW.
To support link from HyperPicture to WWW, we have taken fully advantage of open architecture of HyperPicture [kirste91a]. We introduce a new type of object "HTML", using Xmosaic as its external viewer, or PIM %(Presentation and Interaction Module) as termed in HyperPicture (A native PIM supports both interaction and presentation). A WWW object is represented as a proxy object in HyperPicture which stores only the corresponding URL. When a user traverses a link pointing to a remote WWW document, the session manager will resolve the URL from the proxy object and activate Xmosaic to access and visualize it.
This image shows that a link from a text object of HyperPicture introducing the House of Computer Graphics, Darmstadt to the WWW home page of our institute Computer Graphics Center is traversed.
In another direction, i.e. from WWW to HyperPicture, a usual way is to develop a gateway which translates HyperPicture information structure into HTML on-the-fly. But for HyperPicture users, we have a shortcut, as each type of HyperPicture objects has its own PIM, either native or external. We needn't to translate HyperPicture object to HTML in order that it is able to be visualized by a WWW client, but let the session manager activate the corresponding PIM directly. This means, HyperPicture only needs to understand HTTP, but not HTML. A disadvantage is, HTML documents accessed couldn't be manipulated as other HyperPicture objects.
In constructing the gateway, there are some problems to solve. For example, HyperPicture has more sophisticated hypermedia data model, such as supporting anchoring within objects which has a native PIM. By translating into HTML document, some information of HyperPicture objects may be modified and even lost. For example, anchors within image objects must be translated into anchors referencing the whole object. Also, link information is separately stored in HyperPicture, therefore they must be inserted into the final HTML objects.