"... In PAGE 1: ...ution. Here the AspectJ equivalent is a pointcut and advice. In our previous work, we identified a dozen of such ba- sic building blocks and labeled them crosscutting concern sorts [13]. These sorts (described in Table1 ) can be used on their own, but can also be composed to construct more com- plex designs or features. For example, the crosscutting fea- tures of the well-known Observer design pattern can be char- acterized as a composition of four instances of concern sorts: two role superimpositions (one for the Subject and one for the Observer role), and two consistent behaviors (one for the notification and another for the observer registration).... ..."

"... In PAGE 6: ... In the following, we provide speci cations of the protocols as needed; Appendices provide the speci cations of Rep and Han. For the ease of comprehension, Table2 records the mapping of agents to the roles they adopt in various protocols presented in this paper. The codes in the brackets by the role names in Figure 3 are used as the debtor and creditor names in the commitments.... ..."

"... In PAGE 4: ...Table2 . Some of the papers, such as [BLG00], [MS94] and [RS01], overlap a number of categories and are shown in overlapping gray boxes.... In PAGE 4: ... This is not surprising, as document selection and viewing time, both measures included in Examine Object , are relatively easy to obtain and are available for every object with which a user interacts. Other areas of Table2 contain little or no work, suggesting possible categories of observable behavior to explore. One likely reason for the dearth of literature across the Minimum Scope categories of Segment and Class is that for many systems, the unit with which the user interacts is the object.... In PAGE 4: ... We classify the papers according to the stage they primarily address, but encourage the reader to explore papers from other categories as well. Design Implementation Evaluation Individual [KB01] [KC02] [MBC+00] [BP99] [BLG00] [BPC00] [KSK97] [Lie95] [LYM02] Group [CC02] [CLW+01] [KOR00] [MS94] [RS01] [SZ00] [JFM97] [KMM+97] [CLP00] Table 3: Papers from the Examine Object cell in Table2 , classified by study type. The papers discussed in greater depth in Section 3 have been highlighted.... In PAGE 5: ... None the less, because such work focuses on supporting the individual, we classify it as Individual . While the papers from the Examine Object category of Table2 spread evenly across several of the categories of Table 3, it is evident that little work has focused primarily on the Evaluation category. This is probably because the field is still young, and until now it has been difficult to determine what sort of evaluation test beds would be appropriate.... In PAGE 5: ... This is probably because the field is still young, and until now it has been difficult to determine what sort of evaluation test beds would be appropriate. 3 Examination of Key Papers In this section, we provide a more in depth analysis of several papers that we believe are good representatives of the various different areas of Table2 . Our purpose in examining these papers in more detail is to present the reader with a better idea of how studies of implicit feedback are conducted, how this feedback is typically used and what the key issues and problems in this area are.... ..."

"... In PAGE 2: ... Table1 . Composition properties supported by standard mechanisms The key idea of the approach motivated in this paper is the separation and independent applicability of these notions by providing explicit linguistic means to express them.... ..."

"... In PAGE 12: ... Prolog is designed for natural deduction; using Prolog, it is easy to represent the AMaP instance as a ground fact, the inference rules as Prolog rules, and to solve the black and white hat problems using queries. The implementation of AGENT in Prolog is compact enough to be included as part of this paper ( Table6 , Appendix C). More importantly, the same Prolog program can be used to solve both the black and white hat problems as we illustrate below.... ..."

"... In PAGE 12: ... Prolog is designed for natural deduction; using Prolog, it is easy to represent the AMaP instance as a ground fact, the inference rules as Prolog rules, and to solve the black and white hat problems using queries. The implementation of AGENT in Prolog is compact enough to be included as part of this paper ( Table6 , Appendix C). More importantly, the same Prolog program can be used to solve both the black and white hat problems as we illustrate below.... ..."

"... In PAGE 2: ... If an object exists, then there is at least one name in the namespace that references it. Table1 . Requirements for namespace consistency.... In PAGE 3: ... In either case, the only possible inconsistency due to failures is that the target object is not referenced by any name in the namespace. We claim that handling orphan objects (violation of property 4 in Table1 ) is easier than handling invalid references (violation of property 3 in Table 1). 3.... ..."