With the prosperity and popularity of social tagging communities, developing an enormous amount of user-generated data with modalities has emerged in recent years. Personalized search based on tag-based profiles, an indispensable and prominent way to assist users to access and retrieve their interested resources, has been extensively studied by research communities. In this paper, we revisit the extant profiling approaches to personalized search in collaborative tagging systems. Specifically, we attempt to answer the following research questions: (i) how many tags are sufficient for user and resource profiling? and (ii) under what circumstances should profile enriching/refining techniques be used to promote the effectiveness of personalized search? The result of our experimental studies in a real-world dataset indicate that the rational size of tags for constructing user/resource profiles does exist. This size can also provide us with an insight into when profiles should be enriched or refined for active/inactive users and resources. We believe the findings of this paper can be quite useful for the future applications in social tagging systems such as user interest predictions or resource recommendations.