Large XML on Small Devices: Techniques Developed in the Fuego Core Project

Storage capacity on limited and mobile devices has risen rapidly in recent years. The amount of available processing capacity is, however, still low, as it is limited by battery capacity. In this talk we look at techniques for efficiently processing the increasing amounts of data in XML format that can be fit on a mobile phone. The techniques were developed by the author and Jaakko Kangasharju as part of the Fuego XML stack in context of the Fuego Core project on mobility middleware. The XML processing techniques included in the Fuego XML stack that we present are: processing sequences of XML particles, parser/serializer byte stream access, parsing in a random access manner, using delayed tree structures and building mutability on these, as well as document packaging for synchronization. In many cases, it is unnecessary to construct the tree structure corresponding to the XML document in order to perform a processing task. Instead, the document may be processed as a sequence of XML particles, such as document start, element start, element end, content, and so on. In the Fuego XML stack, the XAS XML parsing and serialization API supports such sequential processing. We have developed an XML efficient differencing application based on sequential XML processing [1]. Oneway of embedding binary data in XMLdocuments is to use Base64encoding, and embed the resulting text as a text node in the XML document. However, this large text object frequently becomes a bottleneck in XML parsing and serialization. This situation can be remedied by providing access to the raw input and output streams of the XML parser and serializer. Care has to be taken to preserve a valid XML context when switching between byteand XML-level processing. In some cases, XML documents are too large to be accessed in a sequential manner, in which case a parser that can be quickly positioned at