Model-based UI Engineering of Advanced and Creativity-based UIs.

Abstract

Η μεταπτυχιακή εργασία εστίασε σε ένα βασικό ζήτημα της μοντελοκεντρικής προσέγγισης για την κατασκευή διεπαφών. Συγκεκριμένα μελέτησε την ανάπτυξη γλωσσών προδιαγραφών ικανές να υποστηρίξουν σύνθετα σενάρια χρήσης και απαιτήσεις που δεν εξυπηρετούνται επαρκώς απο τεχνικές της τρέχουσας τεχνολογικής στάθμισης. Η πολυπλοκότητα του θέματος των προδιαγραφών διεπαφών προκύπτει αφενός απο τις συνεχώς αυξανόμενες αλλαγές που συντελούνται σε επίπεδο συσκευών και περιβαλλόντων χρήσης και αφετέρου απο την ολοένα και μεγαλύτερη προσφορά βιβλιοθηκών εξειδικευμένων βιβλιοθηκών και εφαρμογών απο την ευρύτερη κοινότητα των κατασκευαστών λογισμικού. Είναι λοιπόν προφανής η ανάγκη αξιοποίησης της συλλογικής εμπειρίας και διαθέσιμων πόρων έτσι ώστε η σχεδιαστική κοινότητα να εστιάζει σε κατεξοχήν σχεδιαστικά καθήκοντα, αντι να περιορίζεται από συμβάσεις και περιορισμούς προερχόμενους απο τις ιδιαιτερότητες της εκάστοτε πλατφόρμας και περιβάλλοντος χρήσης.

With the advent of new interaction platforms, novel network-attachable devices, new input/output devices and mobile terminals the prevalent Graphical User Interface (GUI) design language for interacting with software has changed both in ontological scope and symbolic manifestation. Thus, more recent graphical toolkits signify some sort of departure from or enrichment of the original design commitments in an attempt to facilitate richer interactions with more complex software systems. This trend has brought about customized libraries, widgets of various types as well as toolkit-specific techniques for creating custom interaction elements. Examples of this sort of development include Google‘s Material Design, MetroUI by Microsoft but also earlier efforts such the CrossY system[11] which injects new affordances to traditional widgets so that they can be manipulated using crossing interactions instead of clicking, the OrthoZoom system[12] which turns a traditional scrollbar into a powerful multi-scale navigation tool for very large documents, and scent widgets[13]. More importantly, however, efforts such as the above have enabled the development of new toolkits addressing cutting edge issues in 2D graphical interaction. Amongst the most prominent representatives are Jazz[14] and Piccolo[15] and a variety of information visualization toolkits such as prefuse[16] and JGraph (http://www.jgraph.com/). Such a variety has both positive and negative implications. On the positive site, it enables exploitation of interaction facilities available through different toolkits so as to build creative but also more complex user interfaces. Such interactions are typically engineered using toolkit-based programming and custom techniques, which however bind the resulting user interface to a specific platform. As a result, portability, interoperability and scalability become harder, if at all possible. Another shortcoming is the difficulty associated with the integration of these toolkits with other user interface engineering methods such as model-based user interface engineering. Consequently, the benefits of the latter approach are sacrificed. The basic premise of this thesis is that there is benefit to be gained by appropriating the offerings of both perspectives. Specifically, some of the advantages of toolkit programming-based techniques include the capability to build novel interaction facilities by combining radically different interaction object hierarchies. On the other hand, model-based UI engineering offers powerful abstraction mechanisms to frame UI construction as model transformation rather than code manipulation. Therefore, it would be useful to establish bridges between toolkit-based offerings and abstract notations, device-independent mark-up and the mapping of abstract components to platform- specific toolkit libraries so as to appropriate the benefits of each for the benefit of both UI engineers and end users. In light of the above, the proposed work is motivated by some shortcomings characterizing all recent approaches to model-driven UI engineering. These are summarized as follows: - Model-based UI engineering assumes standard native vocabularies, thus it does not support novel or custom widgets built on top of native components - Model-based UI engineering does not support mixed interaction object hierarchies combining interaction components from different toolkit libraries. By relaxing these constraints, the current proposal aims to extend the capabilities of the current generation of model-driven user interface engineering methods by investigating and proposing: - Techniques for modeling the interaction affordances of a designated interaction platform in a manner compliant to a model-driven development framework such as usiXML[17] -New techniques for specifying (as opposed to programming) and modeling interactive behaviors so as to utilize novel interaction affordances which may be implemented through platform-specific custom widgets. -New tools to enhance the capacity of existing model-based development frameworks such as usiXML to utilize (i.e., link to) implemented (non-native) interaction components