Mobile and wireless information is an area on the increase in Europe. Before the year 2000, 30 million people will be using mobile data devices over the GSM network. But to keep mobility, the information display devices must also become more mobile.At the Bonnier Media Lab, the research unit of the Bonnier group, we have tested and evaluated solutions for publishing on handheld devices ("PDAs") using the GSM network as a data carrier. We have also attempted an evaluation of future solutions for mobile information services.
This project has resulted in two things: Current devices are insufficient, and more bandwidth and better compression is needed before this becomes a serious media carrier. Of the devices currently on the market, the Windows CE computers has by far the best performance and the best display.
Since the tests were conducted, the Apple MessagePad has received a significant upgrade, with built-in Internet functions and faster processor. We have not been able to test this
.
Mobile and wireless information is an area on the increase in Europe. Before the year 2000, 30 million people will be using mobile data devices over the GSM network. [Ovum] But to keep mobility, the information display devices must also become more mobile. The Internet slogan "Information on the Internet is available for anyone, anywhere, anywhen" is true only with very large modifications. Currently, access to this information is in practice limited to static devices, e.g laptops. While it is possible to connect them to the Internet over GSM, the combination becomes unwieldy and requires a table. This was the reason we decided to test PDAs as browsing devices, using the GSM network (which is ubiquitous in Europe).
To understand the scope and content of this project, we will give a very brief introduction to the different parts of the project, and also look at some previous projects in Sweden, which related to this.
In this report, we are studying only PDAs that currently support Internet connections, but we will also discuss some upcoming products. Below is a list of those systems investigated, and some background facts.
In 1993 the firs "real" PDA, was launched, the Apple Messagepad 100. It was entirely built around the Newton OS platform. The Newton OS was - and is - an operating system developed especially for user friendliness. The first versions of the operating system were slow and contained several bugs, but in later versions these have been corrected, and the Newton OS is as stable as any other operating system.
The present version is Newton OS version 2.0 and the only machines on the Swedish market to use them is the Apple Messagepad 120/130. These have a 20 MHz RISC-processor, and a black and white screen with 320 by 260 pixels resolution. All input is done with an accompanying stylus, either as handwriting or by clicking with the pen on interaction elements on the screen.
The computer also has a PC-Card (PCMCIA) slot where several peripherals can be installed, such as a memory card, modem, or a GSM-card for mobile telephone interconnection. The software that is delivered with the system is of the type earlier mentioned.
Picture 1. Newton Messagepad 130
In the autumn 1996 a new type of device was introduced on the market: Nokia Communicator 9000. It was - and is - (as of the moment of writing) the first product to contain all functions for efficient, wireless data communications over the GSM network.
Nokia 9000 has a modified Intel 386 processor and uses GEOS 3.0 as the operating system. It has a background-lit LCD screen with 640*200 pixels resolution and can display 8 gray scales. All input is done through the built-in keyboard, which includes diacritical characters (national character sets). Besides the standard calendar and notepad applications the unit is delivered with software to handle SMS (Short Message Service), a way of sending 164 character long messages over the GSM control channel, i.e. without setting up a connection, applications for telefax reception and transmission (a capability that is built into the GSM network), a Telnet terminal, and a WWW browser. The unit is also a mobile telephone (based on the Nokia 2110i) which can be used to place calls. The battery has a wait time of 30 hours and an efficient call time of 2 hours.
Picture 2. Nokia 9000 Communicator
On November 17, 1996 Microsoft launched its new operating system, Windows CE. At the same time Casio launched its new handheld computer, Casio Cassiopeia. Casio Cassiopeia A-10 has a 32-bitars RISC-processor from Hitachi with a clock frequency of 40 MHz. The unit has 2 MB RAM for user installed software and files (extensible to 4 MB), and 4 MB ROM housing operating system and standard software. It has a pressure sensitive LCD screen with a resolution of 480*240 pixels and can display four gray scales. All input is done through the built-in keyboard, or through a stylus. The latter currently only works as a mouse, but support for handwriting recognition is due.
Besides calendar and notepad, the Windows CE computers are delivered with Microsoft PocketWord, Microsoft
PocketExcel, and Microsoft PocketInternetExplorer. There is also a PC-card (PCMCIA) slot where for
instance memory cards (FlashROM), a modem, or a PCMCIA-GSM card for connection to a mobile phone.
Picture 3. Casio Cassiopeia CE
GSM is an abbreviation of "Global System for Mobile communications", and is a digital system for mobile tele- and data communications. It is a European standard, and has been accepted in several other countries in several frequency variations, using the 1900 MHz variation in the USA. Currently, four services are built into the GSM system:
Voice services (normal mobile connections)
Data communications ( 9600 bits per second)
Fax transmissions, using the same channel as data transmissions.
SMS-transmissions. SMS is an abbreviation of "Short Message Service" and is a service to transmit and store brief messages (up to 164 characters) between GSM service modules, e.g. Nokia 9000. An ordinary PC can also send GSM messages, using a modem to dial into the transmission service.
There are many possible pitfalls when connecting portable computers with the Internet over a wireless carrier. These problems can be summarized in the following bullets:
The purpose of this report is to evaluate the possibility of building services based on WWW for PDAs, and
define design parameters based on the available PDAs. We also intend to compare the transmission capacity on
a GSM telephone connected PDA to a PDA connected via a fixed line to the Internet.
The practical testing in this report was done on only PDAs available on the Swedish market in late 1996, and which support connecting to the Internet over GSM. We have not developed any software, only used software available commercially, as freeware, or as shareware. We have not done any server software development, nor any advanced design or development of HTML pages. Microsoft Frontpage was used as both design tool and server.
The preparatory studies were conducted as literature studies. We also took part in several seminars, and attended the launch of Windows CE at Comdex in Las Vegas.
The following points were regarded in the practical testing:
In the GSM network:
In the PDA:
To evaluate if it was possible to use GSM telephones and small PDAs to retrieve information over the World Wide Web, we built a number of test pages which varied in size and composition to represent a selection of features and file sizes representative for what we (admittedly spuriously) perceived as "normal". To connect to the Internet we used, except for the fixed connection, a Nokia 2110i mobile telephone. To connect it to the PDA, we used a Nokia DTP-2 PC-Card (formerly PCMCIA) which was inserted into the PC-Card (formerly PCMCIA) slot on the Newton and the Casio Cassiopeia. The Nokia 9000 Communicator has a built-in Nokia 2110i, which should make it comparable to the other devices. The fixed connection was tested in the computer room of the Institute for Graphics Technology at the Swedish Royal Institute of Technology (KTH) in Stockholm. The fixed line is accessed over an Ethernet, but the average transmission capacity is theoretically 1,8 Mbps.
The provider used for the dial-up connection was Telia Internet Surf, where the maximal transmission capacity is 28800 bits per second. The connections worked most of the time, with an occasional busy-tone. To average performance over time, the connection was tested over a period of a week, at different times of day, with all devices.
The web server, a Microsoft Frontpage version 1.1 server running on Windows NT 3.1, was situated in the Bonnier Media Lab premises. These are connected to the Internet over a T1 (2 Mbps) through an Ethernet. Traffic on this line is not sufficiently large to introduce latencies, but to be on the safe side, we also measured transfer times after working hours. This gave no significant difference in transfer capacity.
Test number one was intended to calculate the real transmission capacity of the Internet connection by transferring a picture of pre-determined size, and time the transfer to calculate the transmission capacity. The transmission was done using HTTP over PPP, which means that not only the link capacity was tested, but also the performance of the browser and protocol implementations. The user is likely to perceive this as a package with the transmission.
Test number two was intended to check how large a picture an HTML-page can hold in conjunction with text, before the transmission becomes untenable. Here, heuristic criteria were used to determine the waiting time. The test was first conducted with text only, and the transfer time of the text file was measured. We then added the picture to the page, and measured the transfer time. After this, we successively increased the picture size and timed the transfer. From this we were able to construct a maximum picture size, by measuring the time for the transfer with and without picture, and compare this with the acceptable waiting time.
Test number three was intended to measure how many pictures you can have on a page before the screen size becomes a limit. This may be relevant in case of maps, for instance. We did this test by downloading four different pictures and measuring the transfer time. We also (admittedly subjectively) checked the picture quality as displayed. The intent was only to get a rough idea of the possible resolution of the pictures of the page. Looking back, we may have designed these pages somewhat differently, but they still give a fairly good idea about how large and complicated pages can be to be read on a PDA.
To be able to evaluate how fast the tested PDAs were we compared them with a Power Macintosh connected to a T1 line over an Ethernet. The information may seem insignificant, but is relevant for users who use a fixed computer as their regular access machine, and unconsciously will compare this to the wireless access.
The setup of the test system and the practical testing took more time than what was expected. The test pages could have been selected in a more appropriate manner, since they occasionally became too large for the PDA. On the other hand, finding the capacity limits was one of the purposes of the test.
To be able to retrieve information from a WWW server over the Internet on a Apple Messagepad 130, you will have to install two pieces of software in the PDA (in subsequent versions corresponding software is pre-installed). One, the Newton Internet Enabler, makes the Message Pad "internet compatible", i.e. installs the protocol stack. The other is a browser, where there currently are two products: NewtsCape and NetHopper. The latter is a commercial product, but does not support graphics as of this writing. The other, NewtsCape, does. It is shareware. Since one intent of the test was to involve pictures, the choice was easy. We installed NewtsCape.
This device is the most mobile, since all Internet technology is already built in, and the mobile telephone integrated in the equipment. To connect to the Internet, it is sufficient to open the device, press the Internet key, type the address, and press Enter. After a while (sometimes, as we see below, quite a long wile) the page is displayed in reasonable quality. The browser is developed by Nokia, and uses HTML 2.0. Navigation is done with arrow keys, which can be somewhat tedious if the page has several links.
Microsoft has clearly intended for the user to use Internet on their Windows CE PDAs, since these are delivered with a slimmed version of the Microsoft Internet Explorer browser. If you have used Windows 95, the user interface is familiar and the software is pre-installed. It is in our opinion even simpler than in Windows 95.
As a reference to the PDAs tested we used a stationary Apple Power Macintosh 8100/100 with a 100 MHz RISC-processor.It had 32 MB RAM, of which 8 MB was dedicated to the browser, in this case Netscape Navigator 3.01 Gold. A 20 inch colour screen was connected, and set to show thousands of colours. The connection to the Internet was over a T1 connected via Ethernet to the computer. Theoretically, this meant that the fixed connection was 200 times faster than the mobile connection. However, theory and practice occasionally differ.
The test files were created in Frontpage 1.1, and served by the Frontpage server.
| File name | Text in kB | Picture in kB | Linked picture file | Total kB |
| pdatest1.htm | 14 | 14 | ||
| pdatest2.htm | 14 | 5 | bonnier.gif | 19 |
| pdatest3.htm | 14 | 3 | otm.gif | 17 |
| pdatest4.htm | 14 | 14 | ClearFusion.gif | 28 |
| pdatest5.htm | 14 | 19 | Vosaic.gif | 33 |
| text5sid.htm | 12 | 12 | ||
| txt10sid.htm | 27 | 27 | ||
| txt15sid.htm | 43 | 43 | ||
| txt20sid.htm | 62 | 62 | ||
| otm.gif | 3 | 3 | ||
| bonnier.gif | 5 | 5 | ||
| ClearFusion.gif | 14 | 14 | ||
| Vosaic.gif | 19 | 19 | ||
| 100k.gif | 78 | 78 |
Time to fetch a page (in seconds)
| Nokia | Casio | Newton | Fixed line | ||
| File name | Test # | ||||
| pdatest1 | 1 | 55 | 55 | 57 | 1 |
| 2 | 66 | 56 | 63 | 1 | |
| 3 | 74 | 66 | 56 | 1 | |
| Average | 65 | 59 | 59 | 1 | |
| pdatest2 | 1 | 100 | 62 | 72 (without picture) | 1 |
| 2 | 85 | 69 | 76 | 1 | |
| 3 | 75 | 76 | 78 | 1 | |
| Average | 87 | 69 | 75 | 1 | |
| pdatest3 | 1 | 44 | 69 | 118 | 1 |
| 2 | 90 | 59 | 117 | 1 | |
| 3 | 79 | 72 | 127 | 1 | |
| Average | 71 | 67 | 121 | 1 | |
| pdatest4 | 1 | 91 | Failed to load | 120 (without picture) | 1 |
| 2 | 94 | Failed to load | 116 | 1 | |
| 3 | 147 | Failed to load | 123 | 1 | |
| Average | 111 | - | 120 | 1 | |
| pdatest5 | 1 | 175 | Failed to load | 132 (without picture) | 1,5 |
| 2 | 160 | Failed to load | 138 | 1,5 | |
| 3 | 164 | Failed to load | 142 | 1,5 | |
| Average | 166 | - | 137 | 1,5 | |
| text5sid | 1 | 72 | 32 | 120 | 1 |
| 2 | 40 | 36 | 124 | 1 | |
| 3 | 45 | 35 | 123 | 1 | |
| Average | 52 | 34 | 122 | 1 | |
| txt10sid | 1 | Failed to load | 55 | 420 | 1 |
| 2 | Failed to load | 82 | 398 | 1 | |
| 3 | Failed to load | 78 | 410 | 1 | |
| Average | - | 72 | 409 | 1 | |
| txt15sid | 1 | Failed to load | 106 | 500 | 2 |
| 2 | Failed to load | 120 | 500 | 2 | |
| 3 | Failed to load | 124 | 500 | 2 | |
| Average | - | 117 | 500 | 2 | |
| txt20sid | 1 | Failed to load | 105 | 500 | 3 |
| 2 | Failed to load | 102 | 500 | 3 | |
| 3 | Failed to load | 104 | 500 | 3 | |
| Average | - | 104 | 500 | 3 |
PDA performance
| Nokia | Casio | Newton | Fixed line | |
| Test 1 average, bits/s | 3114 | 4002 | 3593 | 451634 |
| Test 2 average, seconds/kB | 3,7 | 2,3 | 8,1 | 0,0 |
| Test 3 average, time per kB picture | 4,6 | 3,2 | 3,8 | 0,1 |
General comments
The Casio Cassiopeia was consistently faster, and obtained a consistently higher transmission speed. However, none of the devices is even near the theoretical speed of 9,6 kbps. It may well be that the difference is a difference in software implementation or processor speed. It should be noted that the Casio uses a 32-bit processor and operating system.
Comments: Apple MessagePad 130
The software was very unstable. The picture was sometimes downloaded, sometimes not. The user interface could also do with a re-design, but this is not a major problem. As all the other tested browsers, it supports HTML 2.0.
A serious bug is that when you hang up the mobile telephone, the Message Pad hangs and does a hardware reboot. We were not able to determine whether this error occurred in the Newton Internet Enabler or NewtsCape.
NewtsCape had more problems with the downloading of text than pictures. The maximum text file that can be downloaded to a MessagePad 300 with a mobile telephone connection over GSM is 2000 characters. The maximum file size of pictures is 4 kB.
Comments: Nokia 9000 Communicator
The Nokia browser can handle both text and pictures in an acceptable manner, but when the texts get too long, the entire page is not downloaded. There is no pre-determined page size when this occurs, since it depends on the availability of the cache memory. When the browser signals that it won't download the entire text, the reasonable waiting time is anyhow long past. The browser implementation is very stable, but there are some minor errors, such as the cache being impossible to shout off and that the "history" list links to the wrong pages.
Comments: Casio Cassiopeia
The Casio Cassiopeia was the fastest device in all tests. It handles picture and text acceptably, but when the pages grow too large, they are not downloaded but cut off. However, like on the Nokia 9000, this has no practical consequences. You are long past the pain threshold for page waiting.
Comments: Stationary computer
All the tests took less than one second per page. The screen gives an excellent overview, and navigation with the mouse is easy and familiar.
The tests show clearly that while it is possible to connect to the World Wide Web and display graphics today, the bandwidth of the GSM network and occasionally the computer hardware and software were possible limitations. It is, however, possible to design a set of rules-of-thumb for pages for wireless access over GSM:
[OVUM] Ovum Ltd: Data over GSM: Market development, London 1996
[ETSI] 1995 "European Telecommunications Standardizations and the Information Society, The State of Art 1995".
[GREEN PAPER] 1996 "GREEN PAPER on a common approach in the field of mobile personal communications in the European Union". Document from the EU, at the WWW on: http://www.ispo.cec.be/infosoc/legreg/docs/greenmob.html November.
[Leskinen] Leskinen, Tom: 1996 "GPRS och GSM-The next generation of GSM". Talk at the seminar "Future of mobile services", Network and Telecom trade fair, Stockholm/Älvsjö, September 1996.
[Reichert] Reichert, Frank: Middleware for a New Generation of Mobile Networks: The ACTS OnTheMove Project, Inet 96 conference contribution, June 1996.
[Huber] Huber, Josef : 1996 "UMTS-the Universal Mobile Telecommunications System". Talk at the seminar "Future of mobile services", Network and Telecom trade fair, Stockholm/Älvsjö, September 1996.
[Hjelm] Hjelm, Johan et al: Business information service definition (On the Move Project deliverable D05); unpublished