Annotations in the co-authoring systems are very important because they can serve as communications among collaborators in the context of the artifact being developed.  Most collaborative writing systems lack sophisticated annotation functionalities to facilitate such communications.  Some "advanced" annotation functionalites are mainly for active reading purposes and do not deal with changes occurred in the document. 

Current prototype allows flexible structuring and filtering on annotations, creating "bundles" for task specific purposes.  A user study has shown that bundles are helpful in reviewing the co-authored documents because it supports flexible workflow and context sensitive communications. 

However, users would have to manually create "bundles" - structure of annotations.  therefore, it might be a burden for the annotators.  so, the proposed solution is to allow tagging of 'keywords' on annotations so that annotators or recepients can easily bundle up associated annotations by filtering with tagged keywords.  Since coauthors have shared context of the collaborated document, the collaborative tags should make sense to the co-authors.

Costs and benefits of bundles and the overall complexity that they bring to the co-authoring tasks need to be tested systematically.  Moreover, mechanism of how to manage bundles over a number of iterative reviewing cycles  needs to be discussed and studied.

Unlike taxonomies, tagging is neither exclusive nor hierarchical, which makes it better for organizing items which do not have explicit hierarchical relationships and the boundaries are less clear.  However, due to the uncontrolled nature of tagging, varous problems relating to synonymy, polysemous and basic level variation are caused.  Data analysis on Del.icio.us reveal a great variety in the set of tags used by users, in frequency of use and the description of items in terms of tags.  However, due to the imitation and shared knowledge among users with similar interests, stable patterns emerge in tag proportions.

The data analysis presented in the paper is well done with the visualized graphs.  The use of stochastic urn dyanamics model to explain the stable patterns of tag protortions is very appealing and interesting.  However, such stable patterns can emerge only for extremely popular URIs.  Unless users are (consciously or unconsciously) open to imitate other users and use collective descriptions for tags, tagging systems will benefit community.  However, as indicated by the popularity and success of delicious and other similar systems, it appears that people cooperate or there exists a commonality in tagging practice in the community.

The author discusses and analyzes the uses of tags in web-page organization tools: Del.icio.us (organizing webpages) and Flickr (organizing pictures).  Since there is no explicit systematic guidelines or standards for tagging, various problems are caused such as ambiguous tags (due to different contexts in which a word can be possibly used), different synomic tags for a word and different practices to tag a phrase (because tags usually are only a single word).  Despite these limitations, the tagging approach works well because of its serendipity, its user-centered approach to organizing information and its demand for less cognitive load from users.

The analysis of the strengths and limitations of tags is done well.   However, I disagreed with some points that the author made about community aspects of tagging.  He stated that immediate feedback leads to a form of negotiation of meaning of the terms in the folksonomy among users.  I disagree that asymmetrical communication between users through metadata occurs at all.  In the scenario the author described, after the user finds out that other users do not use the same tag for an item, she still has to choose another tag to change or add (just as she would have done in the traditional network file sharing system, though it might take longer).  So, it apparently still requires extra work from the user to participate in the community.  Therefore, as Grudin said the disparity in work and benefit exists.  However, there is no doubt that folksonomies (with critical mass) can help users better browse and discover more ideas.

 I'd like to study user experiences of a co-authoring tool that support meta-commentaries and flexible organization/filteration of comments.  Specifically, I'm interested to investigate how the workflow will be improved by using such a tool.

 In the paper, the authors claimed that having a shared visual space will facilitate conversational grounding, and hence help collaborative repair tasks performed more efficiently.  In seek of the empirical data to support their claim, the authors have set up an experiment where a pair of subjects is asked to perform a bicycle repair task under three different conditions (co-located, communicated over audio only and communicated over video and audio).

The study results showed that collaborative repair tasks are performed more efficiently in a side-by-side condition than in a remote condition.  Contrary to their hypothesis, distributed pairs in audio-video condition did not perform faster or more efficiently than those in audio only condition (however, the authors argued that it could be the technical issues and the strategies that the subjects used in the experiment that effect the quality of task performance).  Then, the authors claimed that sharing a visual scene as wide as in the side-by-side condition will better support collaborative remote repair.   Moreover, video systems with better support for feedback and gestures will also improve the tasks peformance.

 The authors did a great and thorough job in setting up the study and analyzing the data.  I am convinced by the authors' argument that shared visual space, which faciliates converstational grouding, is essential for collaborative repair.  It is understandable that in remote locations, having shared visual context does not necessarily decrease the amount of utterances and the task completion time.  Due to more detailed information available through the visual channel, more detailed instructions or explanations can result during the tasks.  Whether such detailed information will improve the tasks or will frustrate the users more will depend on the nature and complexity of the tasks performed.

 In the paper, the authors explored the advantages and disadvantages of using video for collaboration.  The authors supported their claims by using the data from their case study, in which they observed a team of five distributed software engineers.  They also suggested design guidelines for effective and successful video conferencing systems, which they derived from the implications of their findings.

 The authors claimed that adding visual channel to audio channel brought several benefits; most of which are enabled through the gestures and facial expressions that are not available in audio channel alone. Examples include expressing understanding through effortless feedback such as nodding one's head, enabling one to sense the others' feelings through facial expressions or gestures and forecast the responses.

The authors also stated the disadvantages of video compared to face-to-face communications.  Due to the lack of eye contact and other attentional limitations in the video system, it makes it diffcult for the users in managing turn taking, controlling the floor and using peripherial cues.  On the other hand, the author mentioned that video has advantages over face-to-face, such as handling interruption and resulting in more efficient conversations.

While the authors were able to include their case study data to support their claims about the benefits or disadvantages of the video conference system, it is the data and methodologies used that become questionable.  The authors mentioned that "although we took many measures of [subjects'] work activity, the data for this paper are based on videotapes of six interactions", without justifying why they chose those particular six interactions.  It is important, yet challenging, that the study data should be as reprentative/generalizable as possible.  In this case, it seems that the authors delibrately chose to represent the data that only support their claims. 

Moreover, the soical issues, which usually have great effects on communication tools like video conferencing systems, were overlooked in the authors' analysis of the data.  Some of the issues are different contexts in which the communication will occur and possible different ways in which the system will be used (e.g.,whether user will devote their attention completely to the visual information provided, or will perform multi-taks).

Though the paper lacks a more thorough analysis of the data and issues surrounding the video conferencing system and the data might not be generalizable, the design gudielines are helpful where applicable.