Personalized Transcoding Proxy - an Approach to Mobile Web Access
Markku Kylnp and Immo Heino
P.O.Box 1203, FIN-02044 VTT, FINLAND
Wireless LANs and mobile network services (GSM Data, HSCSD, GPRS) make possible to use the web basedapplications in mobile environment. Mobile terminals may vary considerably from an ordinary laptop PC:display size, mass storage capacity and CPU processing power will always be more sparse than desktop PCs. Wireless networks will also have at 10 - 100 times less bandwidth and higher latencies than fixed line networks. Available bandwidth and quality of service may also vary dynamically depending on network type.
Because providing and maintaining Intranet/Extranet services already requires considerable resources it is not likely that content providers would support separate content versions for some special handheld devices or for narrow band network use. Some techniques are needed for dynamic content adaptation of various network and hardware combinations in a way that does not force providers to change existing services.
Many proxy services that adapt TCP/IP protocol stack and some service protocols (like HTTP) have been developed . There are also proxies that manipulate presentation of the content depending on mobile device and mobile network constraints ,. Usually these proxies are statically configured to support single specific devices. Our approach has a slight difference since the final goal of our approach is to support personal mobility and to adapt dynamically to changes of user's network environment and devices.
The purpose of our project (APRO ) is to develop and apply techniques to adapt dynamically content of the WWW service taking into account constraints of mobile devices and mobile networks. The prototype of the APRO proxy server makes it possible to use existing WWW servers and WWW browsers without modificationsin wireless Intranet/Extranet environment by hiding heterogeneous network environments and mobile devices from the service.
Filtering and content modification functionality
GSM Data has some annoying characteristics for mobile users like long connection times, a low throughput and a connection time based billing. When higher speed solutions like GPRS are coming, mobile data transfer will still be quite slow and costly compared to fixed networks.
The most significant part of network throughput is splurged to download pictures and images implanted into Web pages - plenty of those are purely aesthetical decorations or even commercial banners. On the other hand images might form visual "landmarks" for navigation to users who only seldom visit on certain pages or those images consist of some valuable information which we aren't fully aware of. So complete image removal isn't a fruitful approach - instead we can use several image related manipulation methods to conserve the bandwidth:
- Reduction of image depth.
- Format modifications. (e.g. GIF to JPEG)
- JPEG Q-parameter modification.
- Image scaling.
- Big image removal.
- Removal of tags and elements (e.g. applet and script elements)
Previous techniques also relieve the problems with PDA's limited size of displays, but some more advanced methods could be applied:
- Structural changes to presentation format.
- Background paging.
- Estimation of the transfer cost and cost visualization to the user.
Methods described above are also partially applicable to Wireless Application Protocol gateways, where HTML pages should be divided into smaller parts(decks and cards) for displays that have even more limited display areas than PDAs.
Our prototype implementation implements most of the functionality that is described above. The implementation is a thread-based Java proxy server. Proxy users are able to configure their profiles using HTML forms. Proxy users are authenticated using proxy authentication.
As future work more emphasis is needed for dynamic adaptation. Also support for CC/PP  and scalability support by using either proxy cluster or personal office workstation proxies should be studied.
Our testshave shown that loading times ofWeb documents can often be cutto one third to one fifth without any significant loss of content. Due to this improvement users have felt more comfortable to browse pages via GSM network connections.
Average users are not eager to configure behaviour of the system by themselves - usually they are more willing to utilize predefined configurations maintained by system administrators. In the future we hope to be able to dynamically support Quality of Service (QoS, in this case the visual quality of pages) according to the network throughtput without manual intervention using some heuristic rules.
On the down side actual test results with PDAs browsers have shown several limitations with current implementations. Either a proxy support was totally absent, a proxy authentication (which is used for a content personalization in our implementation) was missing or some terminals have a very cumbersome procedures to configure a proxy server usage. We also need more efficient content negotiation methods (CC/PP-standardization is ongoing) before we really can talk about effective wireless web use via content transcoding proxy services.
- Mowgli project, Mobile Office Workstations using GSM Links, University of Helsinki, Department of Computer Science, Finland, http://www.cs.helsinki.fi/research/mowgli/
- IBM Transcoding Technology V0.1.2 Beta, http://www-4.ibm.com/software/secureway/transcoder/
- Spyglass Prism, http://www.spyglass.com/solutions/technologies/prism/details.html
- APRO project, VTT Information Technology, Finland, http://www.vtt.fi/tte/projects/apro/
- F. Reynolds, J. Hjelm, S. Dawkins, S. Singhal: Composite Capability/Preference Profiles (CC/PP): A user side framework for content negotiation, W3C Note 27 July 1999, http://www.w3.org/TR/NOTE-CCPP/