However, neither of these systems addresses the two issues of paramount importance: addressing items or pieces of informationindependently of their format, and making this information acces-sible across a users’ multiple devices.
3. PHILOSOPHICALUNDERPINNINGSFORINFORMATION FRAGMENTATION
Fundamentally, human existence is socially mediated within thelife cycle of creation, assimilation, and exchange of information.From the ancient cave drawings to the medieval abacuses to themodern computers, some of the paramount of human endeavorswere towards the conception of artifacts to sustain one part or an-other of this information life-cycle. In the abstract, informationexists as a continuum surrounding the various facets of everydaylife: information about one’s automobile includes particulars frombank payments to insurance policies, to details about maintenanceand service. In other words, information rarely exists atomically;the tax returns include information about the bank loan; the annualbudget, in turn, includes information about the tax return, and soon.As the amount of this information grows, there arises the need tokeep it in a structure that affords retrieval (or for a variety of otherreasons [19,25]). And any structure, by deﬁnition, is an implicitfragmentation mechanism, geared to splitting up information intomanageable units for easy assimilation and exchange. However,humans have learned to naturally cope with this fragmentation intheir physical world using their advanced cognitive, spatial ,and association skills.Two developments disturbed this coping equilibrium: the in-formation explosion in the last few decades (mandating complexarchival structures and strategies which implicitly fragmented theinformation) and the advent of digital technology (with metaphorsand ideas which lacked a vision for the future, ultimately resultingin the information broken into arbitrary classiﬁcations).
The fundamental cause of information fragmentation is still thefact that data exists on a continuum, that is data is implicitly tiedto other data and is often difﬁcult to abstract and structure. Cur-renttechnologicalsolutionsimposeanarbitraryandoftenunnaturalform to it. A classic example of this fact is the ﬁle folder metaphorthat labors under the assumption that one’s data can be unambigu-ously classiﬁed into a hierarchy. This assumption forces the user toput a piece of information in a single location; when that piece of information falls on different points on one’s continuum, an arbi-trary decision should be made to place that information in the bestlocation for that time, thereby fragmenting the information and re-sulting in inaccurate ﬁling (with respect to real understanding onthat information). In other words, this is a problem of ‘keeping’(where should one store this?) in such a way that ﬁnding (wherewas this stored?) would be easy .
4.1 Heterogeneity of Collections
Even though humans think of information in terms of intercon-nected, context-rich, association of experiences, the same are notreplicated in the digital realm. For example, the idea of a contact(say, a friend) in the physical (real) world is often stored or ﬁled (inlong-term memory) as an abstraction of the friend’s name, placeof residence, mental images, shared experiences, mutual friends,and a variety of other attributes. However, when this informationis brought into the digital world, a signiﬁcant amount of detail andcontext is lost in translation. For example, when we save a URLa friend sent as a bookmark, we lose the context of that informa-tion – it was sent by a friend when we were chatting on the 4
of July – stripping the social cues that might be crucial in retrieving orassimilating that information at a later time. Further, when this in-formationisﬁledinthedigitalworld, thevariousattributesgetfrag-mented because of the architecture of current solutions: address inthe address book, shared experiences in photos or videos, commu-nication in emails, and some other attributes in ﬁles or hypertext.Therefore these digital representations cause information fragmen-tation and result in an unnatural ﬁling structure. Solving this prob-lem is one of the holy grails of PIM research, starting from thevision of a Memex  to more recent efforts such as Haystack .
4.2 The Metadata Problem
An important problem results from the fact that different docu-menttypesandcollectionsinvolvedifferentrelevantmetadataﬁelds.For example, a ﬁle has a creation date and a last-accessed date, anemail message has an associated sender, subject and conversationthread, while music ﬁles have ﬁelds such as genres, artist names,and albums. Material documents (i.e. non-digital artifacts) havemany more dimensions of metadata associated with them: for ex-ample, users may remember a book by its cover, a document by itsthickness, or a paper by its dog-eared corners. Such metadata doesnot translate well to the digital domain.When moving an item between collections, or splitting up anitem into multiple collections, useful metadata is often lost. Forexample, whenever a ﬁle received as an email attachment is savedto a local disk, information such as the sender, subject and links toother messages in that conversation thread are not preserved .
4.3 Multiplicity of Applications
The inconsistency among these metadata ﬁelds has led currentdesigns to require independent applications for each such type; forexample, email programs to read emails , browsers to read hy-pertext, etc. Current approaches to information systems designtreatthedifferenceinmetadataformatsasvalidjustiﬁcationforcre-ating independent applications; attempts to separate content orga-nization from presentation, and thus to use different renderers atopthe same browsing framework have met with limited success .Most applications are designed in isolation to support a partic-ular task (for example, an email application is designed to handleemail only ). Because most applications do not take a globalperspective toward the user’s information as a continuum, they of-ten use different locations to save their respective data. This putsthe burden on the user for maintaining different hierarchies for thedifferent task data . From the scenario described above, Janehas to maintain three chemistry folders in different locations forthe three applications she uses to keep her chemistry information,with all the inherent versioning, contextual, and other extraneousoverheads in the process.
4.4 Diversity in Platforms
The problem of information fragmentation is compounded evenmore with the proliferation of different computing platforms suchas laptops, PDAs, and cellphones. Not only does data exist in dif-ferent formats and locations, requiring different applications, thereare now different dimensions to this problem: versioning (
“whereis the latest version of my document?”
), context preservation (
“inwhat state did I leave my information?”
), data ﬂavors (
“can I keepmy information on this platform and not lose its richness?”
), etc.Information of the same type that is stored on different devicesoften ends up as part of different collections. For example, email