How To Subscribe and Unsubscribe is at the end of this note.
Mailing List Trouble? See http://www.langa.com/help.txt
Want an easier-to read formatted HTML version? See http://www.langa.com/whats_new.htm

Please email the LangaList to a friend! (Use this super-fast form !)

The LangaList

18-Apr-99

A Free Email Newsletter from Fred Langa About BrowserTune,
HotSpots, Columns, Tips & Tricks, and Other Activities

In This Issue:
Browser Failure Patterns Revealed!
More BrowserTune News (and ver. b1.6!)
Attention, Closet Geeks
A Better URL For Microsoft Updates
Speaking Of (Win98) Upgrades…
Just for Grins
More!

Over the last few weeks, tens of thousands of people have test-driven the preliminary version of BrowserTune2000. Surprisingly, a significant percentage of browsers fail even the basic portions of BT2K's diagnostic tests, and the pattern that seems to be emerging is this: What I've found so far is that the browsers most likely to have problems are

• Those encumbered with large numbers of add-ons or plug-ins
• "custom" versions of browsers, such as those altered by ISPs
• browsers installed over ISP-altered versions
• browsers installed over beta versions

And these problems plague Microsoft and Netscape browsers about equally: Almost everyone can be affected by these problems!

Running through problems and fixes for each of these would make this email way, way too long, so instead I made it the topic for this week's week-long discussion on the WINDOWS Magazine BBS area.

There, starting around midday (EDT; UT-6) on Monday April 19 1999, I'll run through the problems you probably will encounter with plug-ins and add-ons, with customized versions of browsers, and with browsers installed over customized or beta versions. You may be surprised at what you see!

Get the scoop, and then join in the week-long discussion starting around midday (EST; UT-6) on Monday April 19 1999; you can access the discussion via the front page at http://content.techweb.com/winmag/ .

return to top of page

If you've already run the BT2K demo ( http://www.browsertune.com/bt2kdemo/ ), check the email report you were sent: If you ran any version earlier than b1.5, you haven't seen the latest. And in fact, as of Monday 19-Apr-99, around midday (EST, UT-5), BT2K version b1.6 will be ready!

The throughput tests using three geographically dispersed servers are working fine and show just how variable the Internet can be: it's not at all uncommon for people to get great performance from one of two servers, and dog-awful throughput from the other(s).

Part of that is local: For example, around the time you're reading this note, tens of thousands of other are also reading it, and many will choose to click over to BT2K. When a server gets very, very busy, things can bog down a bit. That's one of the reasons why the BT2K custom email always contains a warning not to rely too much on any single test, and instead to run BT2K at several different times (not just in rapid succession) so you can help eliminate the effects of temporary Internet logjams.

But part of it is the Internet itself. Because Internet data traffic isn't hard-wired, two successive data packets from your PC to a server, or vice versa, can theoretically follow two completely different paths, each with its own set of built-in delays and latencies. The Internet is constantly in flux, and is never exactly the same twice. That makes testing, um, a challenge. 8-)

return to top of page

If you're not interested in how the tests work, just know that in version b1.5 and higher, the throughput tests are highly accurate, and that the manual tests that will be available (as a supplement to the automated tests when BT2K ships) will be even more so.

But if you're a closet geek (and I know I am), you might find this interesting:

The BT2K timing tests run by calling "external" JavaScripts--- JavaScripts stored in a location away from the actual BT2K pages you see. The three external JavaScripts are identical and are made up of long, compression-resistant text strings, in which the text has very few characters that repeat (I created the scripts myself using a random-character generator I wrote). The lack of compressibility means your modem (or ISP) can't artificially reduce the size of the file---every bit will have to be processed. This helps ensure you'll get an accurate reading rather than one artificially inflated by data compression.

The three external JavaScripts are identical, and of a length that's known literally to the exact number of bits. One script resides on a server in New York, one on the WinMag server farm in California, and one on a server in Florida. By timing how long it takes your browser to receive the script, it's easy to divide the known length by the download time to arrive at a bits/second.

But there's a catch: Calling an external JavaScript means your browser has to contact the remote server, open a connection, access the file, then actually download the file, then close the connection. There's more going on than just downloading the file, and each of those other steps takes a little time.

So, before accessing the actual throughput JavaScript on each server, BT2K first times how long it takes to access download a dummy, empty file from the server. Then, immediately after downloading the real test JavaScript from each server, BT2K then accesses and downloads another dummy file from the server. (It uses a different dummy file to ensure the file will be downloaded from the server and not from your browser's cache, which would give incorrect results.) BT2K then averages how long it took to set up and knock-down the connection for the dummy files for each server, and then subtracts that amount of time from the results of the real downloads.

This way, BT2K can automatically minimize the effects of the set-up/knock-down times for each file, and concentrate on the actual throughput times. While this trick of programming (it was fun figuring it out, let me tell you!) minimizes the effects, it can't eliminate them, and that's one of the reasons why BT2K's tests will almost always show lower throughputs than what you'll see when you (say) download software from the web.

Think of the difference between smooth highway driving where you can kick in the cruise control and sail along at a steady speed for long distances: Your mileage goes up. Conversely, in stop-and-go driving, your mileage goes way down because it's inherently less efficient.

By loose analogy, there's something similar going on on the Internet. Downloading a large file is analogous to cruising on a high-speed highway. Downloading a short file is more like stop-and-go driving.

As of BT2K ver. B1.4, the throughput tests were using 9K files, and this proved too small: It was causing too much "stop and go" effect. As of 1.5, I used 25K files, and that worked much better.

As of 1.6, BT2K will use 35K files, and that should be fine for the automated tests. The supplemental manual BT2K tests that will be available at release will also offer longer files for increased accuracy, especially with high-speed connections (cable modems, etc.).

Isn't it weird how much goes into even a seemingly simple thing like a throughput test? No wonder I'm going gray! 8-)

return to top of page

The Microsoft.com site is wonderfully rich, with a ton a free stuff--- but trying to find your way around the site can be a nightmare because it's so big.

A few issues back, I told you about using the pages at http://www.microsoft.com/windows98/downloads/corporate.asp for Win98 and IE downloads. I prefer it to the Update Wizard because the page above lets you more easily save the downloaded files locally. When or if you have to reinstall, you can run the update locally rather than having to go online and re-download the file.

Perhaps I was being too Win98-centric. Reader Ron Klassen points out that the pages at http://www.microsoft.com/windows/downloads/default.asp give you similar access to all the Microsoft OSes and versions, not just Win98.

Thanks, Ron!

return to top of page

You'd have to visit an International House of Pancakes to find more waffles than have been evident in Microsoft's handling of the Win98 upgrade path--- but at last there seems to be a real plan, and it's the focus of my InformationWeek Online column this week. In short form, the Microsoft plan appears to be this:

This fall or winter, there'll be a full release of Windows probably called either "Windows 98, Second Edition" or "Windows 2000 Personal Edition." It will contain all the bug-fixes and patches to date, plus new drivers, a copy of IE5, and a new proxy-like Internet sharing application. It may also include features of the "Easy PC Initiative" which is designed to reduce setup, configuration, and maintenance hassles. This version of Windows will be available via retail CDs for $89, or preinstalled on new systems

Separately, all the patches and bug fixes and the new features will also be available on a nonretail upgrade CD that registered Win98 users will be able to order directly from Microsoft for about $20.

Users who just want the bug fixes and updates (and not the new features) will be able to download them separately from the Windows Update site, or in a single large download that will be called "Service Pack 1." The size of the service pack isn't yet known. Some sources say it will become available "midyear."

What's your take? Will you or your company make use of the upgrades or new versions; if so, which ones? Do the new features justify the $20 cost of the update CD, or is this just a way for Microsoft to try to turn bug-fixing into a profit center? Does the delay in merging the business and consumer version of Windows open the door to Linux, Apple, Be, and the rest; or is the Windows monolith so large and strong that the delay just won't matter much? Join in the week-long discussion starting Wednesday 21-Apr-99 at http://www.informationweek.com/langaletter !

return to top of page

You know SETI, the Search for Extraterrestrial Intelligence, right? Most of it is passive listening for signals from other civilizations, but on several occasions, earth scientists have deliberately beamed messages into the void.

That prompts Hawaiian reader David Keuning to write:

Hi, Fred

Hey, do you think extra-terrestrials think of our SETI project as SPAM? Ever wonder if the first contact from another planet might be "Please remove me from your list..."

 

return to top of page

See you next issue!

Best,

Fred

( fred@langa.com )

(P.S. Please email the LangaList to a friend! Use this super-fast form !)

An easier-to read formatted HTML version is available in the "what's new" section of http://www.langa.com . All past LangaList issues are also available via the same link.

return to top of page

Administrivia:

Why are you getting this newsletter? There are only two ways to get on the list (direct email request or via the WinMag mail list signup page) so if you're getting this newsletter; your name came to me through one of those channels.

SUBSCRIBE (it's free!): Send email to subscribe-langalist@lyris.dundee.net
UNSUBSCRIBE: Send email to unsubscribe-langalist@lyris.dundee.net
LIST TROUBLE? CHANGE OF ADDRESS? HAVE QUESTIONS? NEED HELP? See http://www.langa.com/help.txt

This newsletter is a free service of Langa Consulting LLC and is Copyright © 1999 Langa Consulting LLC. All rights reserved.

return to top of page