[ SEE THE RESULTS OF OUR BROWSER COMP TESTING ]
BROWSERS: A LOOK AT THEIR ROOTS
HTML TAGS AND WINDOWS 95
HTML EXTENSIONS AND WINDOWS 95
IBM OS/2 WARP BROWSER
A LOOK ACROSS PLATFORMS
KEEPING UP WITH THE HTML 3.2
TESTBED
FINAL ANALYSIS
IBM OS/2 WARP BROWSER
[ BY LISA L. SWEET ] On May 7, 1996, the World Wide Web Consortium (W3C) announced a major release of HTML. While HTML 3.2 (code named Wilbur) is still in its draft form, it has become the de facto standard. And while still providing backward compatibility with the existing HTML 2.0 standard, the 3.2 standard adds features such as tables, Java support, heading alignment, bullets and numbers for lists, the copyright symbol, and big and small font tags.
Due to the explosion of Web usage and the outbreak of more sophisticated Web-enabled applications, more and more users are finding their existing browsers obsolete. In addition, webmasters developing pages with the HTML 3.2 standard may not be delivering the content they had in mind. To sum up, users cannot take advantage of many of the new features offered in the HTML 3.2 standard because some browsers simply do not support many of these new features.
ZD Internet Labs used BrowserComp™ 1.0, created by Ziff-Davis Benchmark Operation, to put a series of popular browsers through an extensive HTML 3.2 compatibility test (a copy of BrowserComp™ may be obtained from www.zdbop.com).
Our overall goal was to run eight browsers through their paces on six operating systems to determine how well each one managed standard HTML code. More than 170 HTML tags and 12 extensions were visually inspected for accuracy. Not only did we test the latest versions of Microsoft Internet Explorer, Netscape Navigator, Spyglass Mosaic, NetManage WebSurfer, IBM WebExplorer, and America Online's browsers, but we tested each browser across each platform it was available on, meaning Windows 95, 3.1 and NT, Macintosh, OS/2 and Unix platforms (see chart above for complete information). In some cases, the results even surprised us.

BROWSERS: A LOOK AT THEIR ROOTS : : To understand why the browsers tested the way they did, you have to look at the family tree of browser development. Unlike the other vendors mentioned in this review who are in the browser development business, one of Spyglass's charters is to sell its browser technology to companies developing browsers. In May 1994, Spyglass licensed its first commercially enhanced version of Mosaic. In December 1994, Microsoft signed a license agreement for Spyglass Mosaic and to date, still uses some of Mosaic's code in the Internet Explorer product. Currently, Spyglass licenses its technology to more than 100 partners and recently released Mosaic 3.0.
The NetManage Chameleon product family has been a long-standing citizen in the world of TCP/IP tool suites. NetManage started bundling its WebSurfer browser in its Chameleon suite in the fall of 1994. In November 1995, NetManage started offering its WebSurfer browser free-users no longer needed to purchase the entire suite of tools to use the NetManage browser.
IBM Web Explorer for OS/2 Warp 3.0 was released in October of 1994 and was written in-house. However, the next release of OS/2 Warp 4.0 includes Netscape Navigator as its Web browser.
AOL launched its browser capabilities in May 1995. Version 2.5 of AOL's browser was based on Booklink Technology's Internetworks product. AOL's latest 3.0 version for Windows is an enhanced version of the 2.5 browser. An interesting note is that AOL will be integrating Microsoft Internet Explorer in the fall. A new version 3.0 for the Macintosh is also scheduled for release shortly.
Using its own technology, Netscape made its entry into the browser market in August 1995 with Navigator 1.x. In February 1996, version 2.0 was released, and in August 1996, Netscape released 3.0.
At about the same time Netscape Navigator was first launched, Microsoft announced the initial release of its Internet Explorer browser. In this initial release, Microsoft used a great deal of Spyglass Mosaic code. However, Microsoft wrote much of its own code for Internet Explorer 3.0, which was released on August 13, 1996. However, 10 percent to 20 percent of the code in Explorer is still based on Spyglass Mosaic code. And after looking at our test results, you can readily see the similarities.

HTML TAGS AND WINDOWS 95 : : In reviewing the Windows 95 tests for each browser, we have denoted the general form of each tag as <TAG ATTRIBUTE=VALUE>. Netscape Navigator 3.0 definitely had the highest compatibility success rate. However, one tag that didn't pass the test was the Table Align Center tag <TABLE ALIGN=CENTER>. Incidentally, that attribute failed on every browser and platform with the exception of AOL Macintosh 2.7.
It is important to note that unordered list bullet <UL TYPE=SHAPE> and <LI TYPE=SHAPE> tests passed, where on all other Windows 95 browsers, these tests failed. These tests prove that a browser can display different types of bullets such as round or square in unordered lists. Another test Navigator didn't pass was the definition tag <DFN> tag which should have altered the appearance of text in a document. However, this is fairly insignificant, as the HTML 3.2 specification does not explicitly state how a browser should display text marked with the <DFN> tag.
Our hats are off to Microsoft Internet Explorer 3.0, which had the second highest success rate, although the tests Internet Explorer failed were in slightly more significant areas. For instance, the unordered lists, list level bullet types and item level bullet types (<UL TYPE=SHAPE> and <LI TYPE=SHAPE> ) didn't pass. And in the <BODY> tag, the ALINK attribute failed. When a blue hypertext link was clicked on, it failed to turn red, denoting that it was active.
We found the other browsers to be pretty much "middle of the road" in respect to HTML compatibility. However, the most glaring incompatibility was the lack of Java support in Spyglass Mosaic, NetManage WebSurfer, and AOL's browser. In addition, these browsers didn't support the <DIV ALIGN=CENTER> tag, nor did any of the browsers support the unordered list level bullet types (<UL TYPE=SHAPE>, item level bullet types <LI TYPE=BULLETTYPE>, or the <BODY> ALINK attribute as seen in Microsoft Internet Explorer's testing.
We also found quite a few incompatibilities in respect to text tags such as subscript <SUB> and superscript <SUP> for Mosaic and the complete lack of the capability to change item numbering in ordered lists, <P ALIGN=ALIGNMENT> and <DIV ALIGN=ALIGNMENT> aligned text, proportional and fixed height tables <TABLE HEIGHT=x>, and the <TD NOWRAP> attribute in table cells. In addition to similar incompatibilities as stated above, NetManage had its hands full of other incompatibilities. WebSurfer didn't support color <FONT COLOR=x>, with either named color or hexadecimal specified. In addition, some scrollable text boxes in the forms tests failed, there was no image border support <IMG BORDER=x>, nor did it support Client Side Polygon Image Maps <AREA SHAPE=POLY>. It even proved incompatible with strikethru <STRIKE> and big and small text tags <BIG> and <SMALL>. To sum it up, we found NetManage WebSurfer 6.0 for Windows 95 to be the most incompatible with the HTML 3.2 standard on that platform.

HTML EXTENSIONS AND WINDOWS 95 : : In our tests, we tested 12 HTML extensions ranging from blinking and marquee text, static background images, a variety of GIF types, border color, and frame support. An extension is defined as an HTML tag that is not defined by the current specification, but rather developed and implemented by a particular vendor outside of the specification. It is important to mention up front that blinking text and marquee text features are Microsoft- and Netscape-specific. Microsoft supports marquee text <MARQUEE>, and Netscape supports blinking text <BLINK>. All of the other browsers failed both of these tests. However, in Microsoft or Netscape's case, failing or passing either one of these tests is not a strike against either product. Instead, developers must be aware of what extension each browser supports.
Microsoft Internet Explorer passed most of our extension tests, which is not surprising because a number of the tags we tested were developed by Microsoft. IE came in with one incompatibility, the <BLINK> tag, as mentioned above. Navigator was second in incompatibilities by not passing the Microsoft-centric tags, which included the marquee text <MARQUEE> and the static background image text tag <BODY BACKGROUND =URLBGPROPERTIES=FIXED>, where when scrolling down the screen the background was fixed where the text scrolled. Other Microsoft-centric tag extensions that didn't pass were: table background image tag<TABLE BACKGROUND=URL>, cell background image tag <TD BACKGROUND=URL>, and border color tag. Both Internet Explorer and Navigator did pass the frames test.
While Netmanage WebSurfer 6.0 passed the frames test, like AOL's browser did, we found both to be incompatible with the blinking text tag <BLINK>, marquee text <MARQUEE>, static background image test <BODY BACKGROUND=URL BGPROPERTIES= FIXED>, the animated GIF image test, and the table and cell background image test (<TABLE BACKGROUND=URL> and <TD BACKGROUND=URL>). However, we found that AOL's browser didn't offer support for frames, the trademark symbol tag, or border and background color tests <TABLEBORDER=1BORDERCOLOR =RED> and <TABLE BORDER=1 BGCOLOR=RED>.
The biggest disappointment was Spyglass Mosaic, which only passed two of our extension tests-the trademark symbol and frames tests. However, we'll give Spyglass credit for passing the latter, as the frames function is an important feature.

IBM OS/2 WARP BROWSER : : Because IBM offers its WebExplorer 1.1d with OS/2 Warp, we decided to put it through our tests. WebExplorer was even more incompatible than WebSurfer 6.0. To get an accurate picture, you could take most of WebSurfer's incompatibilities and add fewer supported text styles, no support for any Client Side Image Maps <IMG ISMAP USEMAP=MAPNAME>, no right table alignment <TABLE ALIGN= RIGHT>, and no image size scaling support <IMG WIDTH=x HEIGHT=x>. The frames test even proved incompatible. Well, you get the picture. IBM's current browser is nothing to boast about, but with the bundling of Netscape Navigator with OS/2 Warp 4.0, there is light at the end of the tunnel.

A LOOK ACROSS PLATFORMS : : Just because a browser behaves well on one platform, don't expect it to behave the same way across all platforms. We found Netscape to be the most consistent across all platforms, including our Unix tests. It showed only two to three minor incompatibilities in the HTML tag testing.
However, a trend that we noticed is that the Windows for Workgroups and Mac platforms failed the Java test across all of the products we tested. Microsoft Internet Explorer had the second fewest incompatibilities; however, there was quite a variance across platforms, ranging from five to nine incompatibilities. In summary, the Windows for Workgroups and Mac platforms proved to be the worst performers across all browsers.
We found NetManage WebSurfer to deliver the most inconsistent results with respect to our HTML testing across platforms. Its incompatibilities ranged from 29 on the Windows 95 platform, 38 on the Macintosh platform, and 40 on the Windows 3.x platform. Spyglass Mosaic did a little better with most of its inconsistencies on the Mac platform and had the fewest on its Unix platform. But we did uncover some strange results in our forms testing while running a test that aligned text next to a GIF image <INPUT TYPE= IMAGE ALIGN=TOP>, etc., where the GIF images that we used appeared with scroll bars beside them (see screenshot).

KEEPING UP WITH THE HTML 3.2 : : Although HTML 3.2 is considered the de facto standard for developing new Web applications, the specification is not yet finalized. Those interested in keeping up-to-date with these changes may turn to: http://www.w3.org/pub/ WWW/Mark Up/Wilbur/. This site points to the main menu for all of the information pertaining to the HTML 3.2 standard.

TESTBED : : Our testbed consisted of a 10Mbps Ethernet network with a HP Netserver LH Plus Microsoft NT 4.0 server running Internet Information Server 2.0, and a variety of seven client workstations. Our client workstations were all equipped with 17" monitors and were configured as follows: five Dell Pentium 166 workstations equipped with 32MB of RAM, 2GB hard drives, one Apple 8500/180 system equipped with 32MB of RAM, 2GB hard drive, a Silicon Graphics Indy equipped with 64MB of RAM, 2GB hard drive. The Dell workstations had the following operating systems installed: Two Windows 95 workstations, one Windows NT 4.0 workstation, one Windows for Workgroups workstation, and one OS/2 Warp 3.0 workstation. The Apple was running MAC OS 7.5.3, and the Indy workstation was running IRIX, 6.2.
The following browsers were tested: America Online's browsers (for Windows 95-3.0 and Mac-2.7), IBM WebExplorer (for OS/2-1.1d), Microsoft Internet Explorer (for Windows 95-3.0, Windows 3.1-3.0, Windows NT-3.0, and Mac-2.0), NetManage WebSurfer (for Windows 95-6.0, Windows 3.1-6.0, Windows NT-3.0, and Mac-3.6.1), Netscape Navigator (for Windows 95-3.0, Windows 3.1-3.0, Windows NT-3.0, Mac-3.0, and Unix-3.0) and Spyglass Mosaic (for Windows 95-3.0, Windows 3.1-3.0, Windows NT-3.0, Mac-2.11, and Unix-3.0). Each browser was tested across all platforms that it was available on.
Each browser was run through BrowserComp™, which is a set of HTML pages that allowed us to step through the 3.2 spec and check each browser for compliance. Each test had a description that let us know exactly what we should be seeing. Each test also had a corresponding check box on a printed check-off sheet. If we noticed that the description did not match what we saw, we checked off the corresponding test on the check-off sheet. Once we finished the testing, we knew which tags and attributes in the 3.2 spec weren't rendered properly by each browser.
There is also a section of the test that includes extensions which are inadvertently weighted toward Microsoft, even though Netscape has more actual extensions by far. This section of the test contains only a few extensions and by no means tries to cover them all. More will be added in the future.

FINAL ANALYSIS : : One of the most important findings in the test is how each browser and platform handles the different tags. For instance, because most browsers allow users to change their own screen font on the fly, developers must be aware that users' screens may appear different than intended. In addition, Internet Explorer always shows marquee text and does not display blinking text, where Netscape shows blinking text, but does not display the marquee text. Other important findings are where Navigator shows a black border using the <IMG BORDER=x> tag (with x being the thickness of border in pixels), IE shows a transparent border. While the 3.2 standard does not state how the border should be displayed, and neither is right or wrong, a big black border and a transparent border can certainly make a difference on how a page appears.
Also, when creating pages using the bullet and numbering tags <UL TYPE=BULLETYPES> and <OL TYPE=x> where "x" would start the ordered list using Capital Letters, Small Letters, Roman Numerals, and so forth, note that all browsers do not support this feature, thus leaving a finished HTML page with the same round bullets rather than varied, or incorrect numbering schemes.
Our pick: Due to Navigator's solid HTML support and its consistency across all platforms, Navigator 3.0 is the most compatible browser. Since most companies will likely always have mixed OS environments and multiple platforms, Navigator 3.0 is the best fit for corporate environments.
For our complete testing results, see www.zdimag.com/content/zdlabs/browser1.html

Email: zdimag@zd.com
www.zdimag.com


Copyright (c) 1996 ZD NET. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of ZD NET is prohibited. ZD Internet Magazine, ZD Internet Magazine MegaSite, ZD MegaSource and their respective logos are trademarks of Ziff-Davis Publishing.