#!/bin/bash
# D.Kovalov, 2008

# Non-debian patched vpns dos not support Target Nets config option. 
# This script would emulate it: resets default routing after vpnc starts 
# and sets DNS after startr/stop of VPNC

#
# User setting:
# What nets are routced through vpnc tun deviceand what's gateway for it
TARGET_NETS="10.12 172.18 192.168.2"
GW=10.12.34.56
# ------------------------------------------------------------------------
# 
#
[ $(uname -s) == 'Darwin' ] || { echo 'This script is for MacOSX'; exit 1; }
[ $(id -u) == 0 ] || { echo Must be root; exit 2; }
# ------------------------------------------------------------------------

PRIMARY_IF=$(echo 'show State:/Network/Global/IPv4' | scutil | awk -F: '/PrimaryService/ {print $2}' | sed 's/ //g')
# ------------------------------------------------------------------------
restore_dns () {
	# try to stop vpnc and ignore error
	/opt/local/sbin/vpnc-disconnect 2> /dev/null 1> /dev/null
	( cat <<-EOF
		# ------------------------------------------------------------------------
		#  restore DNS setting
		d.init
		get State:/Network/Service/(saved)/DNS
		set State:/Network/Service/${PRIMARY_IF}/DNS
		remove State:/Network/Service/(saved)/DNS
	EOF
	) | scutil

}
function stop_me () {
	/opt/local/sbin/vpnc-disconnect || { echo "Can not stop VPNC session. exiting...."; exit 6; }
	restore_dns
}
# ------------------------------------------------------------------------
function start_me () {

	DEFROUTE=$(netstat -nr | awk '/^default/ {print $2}')

	/opt/local/sbin/vpnc || { echo "Can not start VPNC session. exiting...."; exit 3; }

	# ------------------------------------------------------------------------
	route delete default
	route add default ${DEFROUTE}
	for NET in ${TARGET_NETS};do
		route add -net ${NET} ${GW}
	done
	# ------------------------------------------------------------------------
	# let vpnc manage /etc/resolv.conf and get all info we need from it...
	#
	grep '@VPNC_GENERATED@' /etc/resolv.conf
	if [ $? != 0 ]; then
	  echo "/etc/resolv.conf is not @VPNC_GENERATED@"
	  exit 10
	fi
	# ------------------------------------------------------------------------
	DOMAIN=$(grep -v \# /etc/resolv.conf | awk '/domain/ {print $2}')
	NS=$(echo $(grep -v \# /etc/resolv.conf | awk '/nameserver/ {print $2}'))
	# ------------------------------------------------------------------------

	(cat <<-EOF
		# ------------------------------------------------------------------------
		# Save DNS state
		#
		d.init
		get State:/Network/Service/${PRIMARY_IF}/DNS
		set State:/Network/Service/(saved)/DNS
		d.init
		d.add ServerAddresses * ${NS}
		d.add DomainName ${DOMAIN}
		set State:/Network/Service/${PRIMARY_IF}/DNS
	EOF
	) | scutil

}
# ------------------------------------------------------------------------
case $1 in 
	"start")
		start_me
	;;
	"stop")
		stop_me
	;;
	"restore")
		restore_dns
	;;
	*)
		echo "Usage $0 (start|stop|restore)"
		exit 5
	;;
esac

---

The saga of the computing industry is rich with outsize characters and surprising plot turns, but there's one story that has risen over time to mythic proportions. It's the tale of how software pioneer Gary Kildall missed out on the opportunity to supply IBM (IBM ) with the operating system for its first PC — essentially handing the chance of a lifetime, and control of tech's future, to rival Bill Gates and Microsoft Corp. (MSFT ). In the process, he may have missed out on becoming the world's richest man. The legend goes like this: One fateful day in the summer of 1980, three buttoned-down IBMers called on a band of hippie programmers at Digital Research Inc. located in Pacific Grove, Calif. They hoped to discuss licensing DRI's industry-leading operating system, CP/M. Instead, DRI founder Gary Kildall blew off IBM to gallivant around in his airplane, and the frustrated IBMers turned to Gates for their operating system. This anecdote has been told so often that techies need only be reminded of "the day Gary Kildall went flying" to recall the rest. While he's revered for his technical innovations, many believe Kildall made one of the biggest mistakes in the history of commerce. But what if that's not what happened? What if IBM and Microsoft deprived Kildall not only of untold riches but also of the credit for a seminal role in the PC revolution? That's the thesis of a chapter about Kildall in They Made America, a serious coffee table history book by renowned author and former newspaper editor Harold Evans. The book, published by Little Brown on Oct. 12, profiles 70 American innovators and is the inspiration for an upcoming PBS series. And while other tech authors have debunked the gallivanting story before, Evans bases his Kildall chapter on a 226-page, never-published memoir written by Kildall just before his death in 1994. Early on, Kildall seemed to represent the best hopes of the nascent computer industry. But by the time he died at age 52, after falling in a tavern, he had become embittered and struggled with alcohol. They Made America is certain to elicit cries of protest. That's because it attacks the reputations of some of the key players of the early PC era — Gates, IBM, and Tim Paterson, the Seattle programmer who wrote an operating system, QDOS, based partly on CP/M that became Microsoft's DOS. Evans asserts that Paterson copied parts of CP/M and that IBM tricked Kildall. Because Gates rather than the more innovative Kildall prevailed, according to the book, the world's PC users endured "more than a decade of crashes with incalculable economic cost in lost data and lost opportunities." David G. Lefer, one of Evans' two collaborators, says: "We're trying to set the record straight. Gates didn't invent the PC operating system, and any history that says he did is wrong." There's no doubt that Kildall was one of the pioneers of the industry. He invented the first operating system for microcomputers in the early 1970s, making it possible for hobbyists and companies to build the first personal computers. Legalities aside, Microsoft's original DOS was based in part on Kildall's CP/M. His insight was that by creating an operating system separate from the hardware, applications could run on computers that were made by different manufacturers. "What really drove Gary was inventing things," says friend and former DRI executive Tom Rolander in an interview with BusinessWeek. Still, Evans' book falls short of clarifying exactly how Kildall lost out to Gates. He relies primarily on Kildall's memoir, his family, and his friends. Evans says he requested an interview with Gates, which he says Microsoft denied. He didn't make contact with IBM or Paterson, but tapped previously published accounts for that side of the story. IBM would not talk to BusinessWeek for this article, but former IBMers take issue with Kildall's version of events. Microsoft calls the book "one-sided and inaccurate," and says the company is proud of the "foundational role" it played in the industry. Paterson denies he stole Kildall's intellectual property. He says he's stunned that the authors failed to get in touch with him. "You'd think they might have asked. I'm not hard to find," he says. HAZY MEMORIES What's hard to find is the truth. A dozen interviews by BusinessWeek with people on all sides paint a blurry picture of those crucial days in the summer of 1980. While Kildall claims in his memoir that he met with IBM that first day and reached a handshake agreement, DRI's own lawyer at the time, Gerry Davis, says there was no deal. One of the IBMers who visited DRI that day insists he didn't talk to Kildall, but another, Jack Sams, now retired, says it's possible he was introduced to Kildall, although he doesn't remember it. Sams says faulty memories and self-serving accounts make it nearly impossible to tell exactly what happened during those chaotic weeks. "Back in those days, there was a lot of misinformation that was deliberate," he says, pointing out that IBM originally claimed it had made the PC all by itself. "We spun it, Kildall spun it, and Microsoft spun it." The story begins unambiguously. A group of IBMers, working on a secret project to build a personal computer, flew to Seattle in August, 1980, to see if Gates could supply them with an operating system. He couldn't — and referred them to Kildall. When they showed up at DRI's offices the next day, Kildall's then-wife, Dorothy McEwen, the company's business manager, refused to sign their nondisclosure agreement. She is now ill with brain cancer and can't remember the events, according to daughter Kristin Kildall. But Rolander, who flew with Kildall on a business trip that morning, tells BusinessWeek they returned in the afternoon and Kildall did meet with IBM. If Kildall struck a handshake deal that day, it didn't stick. Sams says he did get together with Kildall in Pacific Grove a short time later, but they couldn't reach an agreement. At around the same time, he saw Gates again. He and Gates both knew of the operating system Paterson had built at Seattle Computer Co. As Sams recounts, "Gates said: 'Do you want to get QDOS, or do you want me to?' I said: 'By all means, you get it."' Gates bought Paterson's program, called QDOS, for $50,000, renamed it DOS, improved it, and licensed it to IBM for a low per-copy royalty fee.

THE SHOUTING It wasn't until nearly a year later that Kildall discovered that Gates, a longtime friend, had plucked the plum software deal out of his grasp. IBM sent test versions of its PC out shortly before it was announced in August, 1981, and a consultant working for DRI noticed the operating system was remarkably similar to CP/M. The consultant, Andy Johnson-Laird, remembers that Kildall looked at the screen and was stunned. "There were some shallow changes, but it was essentially the same program," says Johnson-Laird in an interview with BusinessWeek. Kildall was furious. He and DRI's vice-president for marketing, John Katsaros, met with Gates in a Seattle restaurant to hash things out. "It was one of those meetings where everybody was nice to each other, then everyone shouted at each other, then everyone was nice to each other, then everyone shouted at each other," recalls Katsaros in a BusinessWeek interview. Nothing was resolved. Kildall also confronted IBM. But his problem was that software copyright had just become law three years earlier, and it wasn't clear what constituted infringement. Davis, the DRI lawyer, believes that based on the number of similarities DRI's forensic consultants found between the original DOS and CP/M, "in today's world, you could take it to court and get an infringement." But not in 1981. So rather than sue, Kildall agreed to license CP/M to Big Blue. He was floored when the PC was released and IBM charged $240 per copy for CP/M and just $40 for DOS. Kildall's conclusion, according to his memoir: "I believe the entire scenario was contrived by IBM to garner the existing standard at almost no cost." Within a couple of years, the IBM PC was the undisputed champ, and Microsoft was the leading operating system provider and on its way toward PC industry domination. CP/M gradually faded into irrelevance. Kildall ultimately sold his company to Novell Inc. (NOVL ) in 1991 for $120 million. He went on to create some pioneering multimedia technology, but never again was an industry player. Friends say that, for years, he cringed when people brought up the "flying when IBM visited" story. The last straw was when the University of Washington in 1992 invited Kildall to attend the 25th anniversary of its computer science program. He was one of its earliest and most distinguished graduates, earning a PhD, yet they had picked as keynote speaker Gates, a Harvard dropout. Kildall says it was this dig that prompted him to write his memoir. "Well, it seems to me that he did have an education to get there. It happened to be mine, not his," Kildall wrote. Kildall's resentment is understandable, but even his friends agree that he was partly to blame. For all his technical brilliance, he was a poor businessman. One big mistake was not moving ahead fast enough with a more advanced version of CP/M. He was slow to deliver a 16-bit operating system. It was that delay that created an opening for Paterson to design a 16-bit alternative, and because DRI didn't have its own version ready in the summer of 1980 IBM decided to deal with Gates, says Sams. Once IBM agreed to market his software, Kildall demanded a relatively high royalty — contributing to its being priced so high, say former DRI execs. Would history have taken a different path if Kildall triumphed in those early days? "I'm convinced," says John Wharton, a tech consultant and Kildall pal. He believes the industry would have been more collegial and innovative if Kildall rather than Gates sat at the crossroads of computing. But others say Kildall didn't have what it took to lead an industry. "Bill succeeded because he was a tenacious businessman," says lawyer Davis. "Gary was not tenacious." As for Kildall's family, they're grateful his story is finally being told. "The truth is different for everybody," says daughter Kristin. "I think everybody believes they're presenting the truth. Obviously, they're different. I don't know why. I'm just glad my truth is out there." History may typically be written by the victors. But in this case, Gary Kildall has secured — and deserves — more than just a footnote.

By Steve Hamm in New York and Jay Greene in Seattle

Псевдо: iip
Ім'я: iip
Розташування: Новачок
Дата реєстрації: 17. серпня 2004 17:24
Ел. адреса: схована

Note: This documentation is for Tiki, version 1.8. It probably works for earlier versions, but I haven't tested it. For Tiki version 1.9 and later there is a different approach that requires editing db/tiki-db.php and using some new features in tiki-install.php. The newer variant is scantily documented at http://tikiwiki.org/MultiTiki19.

• How to Host Multiple TikiWikis on a Single RedHat System.

by George Geller?

• №1
  • Introduction
  • MySQL Setup
  • Apache Configuration
    • DocumentRoot (optional)
    • VirtualHost
  • Install TikiWiki
    • Install Distribution
    • Add Database Selection Code
    • Set Up files
    • Run Tiki Install Script
    • Admin TikiWikis
• №2

---

Introduction

Here I show one way to install multiple tikis on a RedHat Linux server
when you have root access. I assume that you already have apache, php
and mysql all working on your system. If you installed the correct
RedHat packages and didn't change the configuration files too much,
these programs should be working. I've had success with this approach
on RedHat 8, RedHat 9 and Yellow Dog 3.0.

In this example I go through the steps to create threee working tikis on
my machine and give a few pointers in the process. I will be
specific, rather than general, concrete, rather than abstract. If
your setup is different you'll have to adapt the instructions.

I illustrate how to set up three tikis: public.wikiplanet.com,
member.wikiplanet.com and staff.wikiplanet.com.

---

MySQL Setup

Each tiki instance will need a separate database in MySql Tiki also
needs a MySql user that has the correct permissions to manipulate the
databases. The names of the databases don't matter to MySql, but I
have adopted a convention where I always use a prefix of tiki_ for the
database name. So for my three tikis, the corresponding databases
will be: tiki_public, tiki_member and tiki_staff. I create a single
MySql user, 'tiki', that will have access to all tiki databases.

The commands I issue as root are:

Then I run MySql to set up the 'tiki' user with the correct permissions

---

Apache Configuration

Next I edit the Apache web server configuration /etc/httpd/conf/httpd.conf so that apache will know where to look for the tikis.

DocumentRoot (optional)

I changed DocumentRoot to /www because I am a lazy typist. The default is /var/www/html.
There are two lines in httpd.conf that you will need to change:
DocumentRoot "/www"
and 25 lines further down:

You can change them, or you leave them and adjust the rest of these instructions accordingly.

VirtualHost

Further down, find the line with NameVirtualHost and change it to:
NameVirtualHost*:80

At the end of the httpd.conf, add the line:



Create a .vh file in /etc/httpd/conf.d/ for each of your virtual hosts.

/etc/httpd/conf.d/public.wikiplanet.com.vh contains:


/etc/httpd/conf.d/member.wikiplanet.com.vh contains:


/etc/httpd/conf.d/staff.wikiplanet.com.vh contains:


Restart your web server.

Don't worry about the warnings that the directories don't exist, we're going to make them next.

---

Install TikiWiki

Install Distribution

Get a distribution of TikiWiki from sourceforge.net. Unpack it and make the soft links.
I used tikiwiki-1.8.RC2.tar.bz2.

Add Database Selection Code

Now you need to edit the code in db/local.php.

# cd /www/tikiwiki-1.8.RC2/db{CODE}
Edit local.php. Mine reads:
{CODE()} 

Set Up files

Set up the files and directories for tiki:

Run Tiki Install Script

Next we need to run the Tiki install script for the new wikis with a web browser. You must have dns working correctly, or you can edit /etc/hosts. In my case dns works, so I point
Mozilla to: http://public.wikiplanet.com/tiki-install.php and run the install script. Do not choose to "remove" tiki-install.php yet! Repeat the procedure for the other two tiki's: member and staff.
In the final case, "remove" the install script.

Admin TikiWikis

That's all. Login to the tikis as admin and configure them.

This page describes how to setup multiple tiki sites using one install but NOT using Virtual Hosting.

by Akira?

In my example I am using Red Hat 7.3, apache 1.3.27-4, php-4.1.2-7.36, mysql-3.23.58-1.73.

The key to this example is that I want multiple sites but only under one hostname.

First setup your databases. I have chosen to create a seperate mysql user and password for each database. You can do this or have a common user/password. If you have a common user/password you will need to change the code for local.php below. Make sure you follow all the steps for creating the databases as outlined in the INSTALL file. Also outlined here: RecipeMultiTiki. Don't forget to restart mysql.

I plan to put my single tikiwiki install in /var/www/tikiwiki/tikiwiki. Knowing this I setup my httpd.conf. I am going to use three subdirectories under my one hostname using aliases. Here is what one looks like, repeat however many times you want changing only the Alias line:


Restart your webserver.

Get a distribution of Tikiwiki from sourceforge.net. Unpack it and place it where you want, creating directories if you haven't already done so.


Next we create the databases through the web install script. Load http://your-domain.com/tiki-test1/tiki-install.php. First fill in the first database and account information. Submit and go to the next screen. Chose the type of database you want to create or update and submit the form. Next click "Execute the Tiki installer again". Then click "Reset database connection settings" and enter your next install's database information. Rinse and repeat these steps for as many sites as you are creating. Once you have created the last site click "Rename/remove tiki install script and proceed to site."

We need to run setup.sh to setup the files and directories for tiki as well as for the multiple sites. Replace tiki-test1, tiki-test2 and tiki-test3 with the alias names you used above in httpd.conf.

Now we want to edit db/local.php. The custom php below does what I want. It extracts the web path from the variable "$SCRIPT_URL". For each site you will need a "case" stanza. The case stanza should include database user/pass/name information as well as the tikidomain variable which we chose above when running setup.sh. If you want to have one mysql user/pass for all sites you would move the $user_tiki and $pass_tiki of the switch section, placing it just below $host_tiki.

$db_tiki     = 'mysql';
$dbversion_tiki = '1.8';
$host_tiki   = 'localhost';  
 
switch ( extractTopPath( $SCRIPT_URL ) ) {
 case '/tiki-test1/':
  $user_tiki = 'tiki-test1';
  $pass_tiki = 'foobar1';
  $dbs_tiki = 'tiki-test1';
  $tikidomain = 'tiki-test1';
  break;
 case '/tiki-test2/':
  $user_tiki = 'tiki-test2';
  $pass_tiki = 'foobar2';
  $dbs_tiki = 'tiki-test2';
  $tikidomain = 'tiki-test2';
  break;
 default:
  exit();
}

function extractTopPath( $in ) {
        if ( ( $point = strpos( substr( $in, 1 ), "/" ) ) === false ) return ( $in . "/" );
        return substr( $in, 0, $point + 2 );
}

Now you are ready to configure each site! Since tikiwiki uses cookies we have to make sure the cookies know which site you are at. To do this you need to login as admin and go to the Login administration page. Toward the bottom is HTTP(S) server name, HTTP(S) port and HTTP(S) URL prefix. You will want to fill the approiate fields in. For my example site I would fill these fields in like this:

HTTP server name	domain.com
HTTP port	80
HTTP URL prefix	tiki-test1/

If you are going to use the remember me feature you will also want to fill out these fields: Remember me domain, Remember me path.
For my example site I would fill these fields in like this:

Remember me domain	domain.com
Remember me path	tiki-test1/

After that you should be able to login to each site and configure them and everything will be seperated!

If you have any questions or suggestions please leave them in the comments of this page.

-Akira

If Nielsen's status today is less Papal, he's nevertheless continued to build a body of research, best practices and technologies (including a remarkable 73 patents), all of which relate to how companies can actually get the increased productivity and business value from IT that has so long been promised. "It's still a horrible experience to do business with most companies," Nielsen says, "because, honestly, their computer systems remain so cumbersome and customer-hostile."

In 1998, Nielsen and fellow futurist Don Norman formed the Nielsen Norman Group. The Group has consulted on projects at a number of well-known firms, including General Electric, General Motors, UPS, Hallmark Cards, and wsj.com. Each year, it recognizes the "Ten Best Intranets." Meanwhile, in his regular online column, "Alertbox," Nielsen continues to increase his reputation as a gadfly.

Executive Editor Brad Wieners recently caught up with Nielsen by telephone and asked him what, if anything, we've learned in the decade since everyone awoke to the need for a Net strategy.


CIO Insight: Let's start with your research findings, some of which seem rather sensational. You report that an average mid-size company can expect a return on investment of 1,000 percent, and a gain of $5 million a year in employee productivity, simply by improving the usability of its intranet.

Nielsen: That's right.

How do you figure?

That number came from 2002. And I am sorry to say that I've not seen huge improvements in intranets these last two years, so that potential is still there. But I also want to point out that that estimate is really going from a company being average to above average, to just being in the top 25 percent. It's not going from average to being the best, which would have even more of an impact. And that particular figure million per year is estimated for a company of 10,000 employees. But this is linear, so it applies to companies of all sizes. Multiply the number of employees by the number of hours per year that they waste.

Then there's this even more eye-popping number: that superior intranet design could save the world economy $1.3 trillion.

That's true, but that's really, truly worldwide. And that's how I can say that usability is not just a small issue. It's one of the biggest driving factors for really getting our productivity up in the white-collar economy, the service economy, and it's really in many ways the equivalent of what was done in the old days when people were studying productivity on the assembly lines. The problem is that nowadays most work is knowledge work. So this means to get productivity gains today we have to adjust the machines—and by machines now, we really mean software. So how do we adjust machines to human thinking? Well, by studying human thinking; in other words, by doing these usability studies and adjusting the technology.

In 1999 you revisited your "Top Ten Mistakes in Web Design" from 1996. Let's compare 1999 to now. What continue to be the most persistent challenges?

Bad search continues to be a problem today even though, from a technology perspective, great progress has been made. You can see this plainly when you use the public search engines. They're much better today than they were ten years ago. But the search on individual Web sites or inside intranets is, typically, still bad. And it's bad in all the different aspects of search. It's usually not unified search—no one search can search everything. This is a particular intranet problem. Things are divided up into different knowledge bases, so you've got to know where to search, and if you need to know where to go to search, then that defeats the entire idea.


The other problem about search is the content, which is to say the individual pages, or units of information, are typically poorly described in terms of things like the headline and the summaries, which is all people have to choose from when they get the search-results listing. So if there was just one thing we could fix on the Web, and for intranets as well, I would say let's fix search; that's still the number one single thing that's causing people problems.

What else?

The second thing that's causing the most problems is information architecture, which continues to be driven more by how the information is produced than by how it's consumed. Intranets are usually divided up by which department does which things, as opposed to what tasks employees have, or which work activities people have.


And I'll just mention one glaring mistake that most companies make: They divide up their networks or Web sites between products and supplies and service. There are typically three different places because there are three different divisions doing it. For a customer, however, if I have a certain copier, let's say the X17 copier, and I want toner for that machine, or I want to get it serviced—well, what I want is to go and find my copier and, once I find it, I want to get supplies for my copier, I want to get some trouble-shooting, self-service information. But it's a major effort because these are in different places. So that's something we find almost every time we do a study: that information is not structured in the way that people think of it. And that has been a problem for all ten years.


And then the last thing—I mean, there's millions of these things. But another one I want to mention is lack of clarity in the content. In other words, the descriptions, the actual information, doesn't clearly answer the questions people have. It's all kind of buried under a huge, thick layer of marketing, you know, of hype, and it's not concrete. The content does not explicitly say what you want to know.


Let me give you a very small example. I was looking at a hotel, and so the hotel Web site says, "Ample parking is available," but you have to pay for it, in this parking garage located in the same block as the hotel. Well, that's all very nice, but can't you just tell me how many dollars a day it is to park there? Okay, it's downtown. I've got to pay for parking. I can accept that. But how much is it?

So you definitely advise publishing prices online.

That's the number one specific thing people always ask for, and I think it's a completely mistaken idea that you're going to lose customers if you tell them what it's going to cost, because nobody's going to buy anything without knowing what it costs.

If usability offers so much potential, why does it seem like there's still resistance to doing usability reviews?

There is resistance. It comes in many different ways, and often it's from the highest-level executives who don't even realize how bad their intranet is, because they don't use it themselves.

I'll just pose a very simple challenge to your readers: When is the last time you did your own expense reports, as opposed to handing it off to an assistant to do it? If you had to find out, you know, the rules for getting reimbursed for taking a client out for lunch, and how to actually take the receipt from the restaurant and enter it into the system—it may involve many steps. Well, most executives never actually do those things, and so, therefore, they don't realize that many of these steps can be quite cumbersome. So you really have to go on the intranet, not just to check it out, "Does it look good?," but to actually accomplish a specific task. So that's one really big issue right there.

Then a second issue is that the people who work in IT departments, even if they do try these things themselves, they don't necessarily recognize how difficult the processes are for mainstream employees. For someone in the IT department, the intranet may make a process much, much easier than it is for, let's say, an attorney in the legal department who might have a very high hourly rate, but who is not particularly a computer person.

The third barrier is that even when it is accepted that something needs to be done to improve usability, it often falters on more political barriers. Within the organization, it's not always clear who is in charge of the intranet and can say, okay, we're going to make it this way. Very often, it's a very distributed responsibility, and so a lot of people have to be brought on board, and they might drag their feet, or say that's not their problem this year. And if you have to get 50 departments to do it, it never happens.

It becomes a major coordination problem.

Exactly.

Last fall, you published your most recent survey of "Best Intranets." Any surprises from that research?

You might say that the lack of a clear technology platform was in some ways a surprise because you read so much about this and that solution being supposedly the way to great intranets. In fact, when we go and talk to those companies that have done great intranets—first of all, they all use something different, and, second, all of them say of whatever technology they happen to be using, "Well, we had to make a lot of changes ourselves to make it really work for us." So I think there is a big contrast between advertising and reality, and that these technologies are not all there yet. You really have to take responsibility yourself if you want to get a good solution.


Another thing that was quite striking was that several of these best intranets had reduction of e-mail as being one of their priorities in their project, finding ways of taking information away from e-mail and sticking it into a more kind of organized and searchable space on the intranet. It was probably because you can provide better features on the intranet than you can an e-mail reader, but it's also partly because people are just getting so buried in their e-mail, we've got to take things out of e-mail if at all possible.

The goal used to be the opposite: to put things into e-mail.

That's right, and people have done it too much. Now we've got to retract.

Among your Web site pet peeves, you inveigh against pop-up ads, which you once compared to selling a vacuum cleaner by first dashing someone's ashtray on the floor.

That's completely what it is like, yes. My screen is really precious to me. It's mine; don't go and pollute it.

Okay, but what have you got against PDFs?

Well, the problem is that PDF documents are just not very suitable for online access because they are optimized for print, and they're big linear documents, and, therefore, they're not very good for search. So if you find something that's in a PDF file, it's probably on page 217 or something, and being dumped at page one doesn't really help you that much. And so often you'll miss the information even though it is, in fact, in the file.

Also, the formatting is optimized for print, so it's simply a nice brochure. It's typically letter-sized, and you kind of have to scroll it too much or the type becomes too small and hard to read. And the very first time you experience this, you don't even see the document. All you see is "Now we're loading Acrobat." So it becomes an extra delay that people hate as well.

You're also not a fan of drop-down menus. For example, you'd prefer to type in a two-letter abbreviation for your state, CA, than scroll down looking for it.

Exactly. Because it's much faster and it's less error prone.


The reason I think that drop-downs are so common is that the programmers want to avoid having to validate the input, but it's not really that difficult to write a little routine that checks that you have one of the authorized abbreviations. And it's actually much less error prone because what very often happens is that people who want to enter "California" will end up with "Alabama" because the menu kind of first goes to C, but then it goes back to A. This is a minor irritant, but it's an example of a more general issue which is, Where do you put the burden? Do you put the burden on the computer or on the user?


Another thing that this points to is the general principle that if a task is keyboard-centric, stay at the keyboard as much as possible. If it's mouse-centric, stay at the mouse as much as possible.

Many of our readers have been through a generation or two of their own Web sites and intranets. Any mistakes to avoid when redesigning that are distinct from the original development process?

Well, the biggest difference is that if you have a redesign, you already have a design by definition, and, therefore, I would very strongly recommend starting by not actually doing anything new, but by researching what you already have. This is a piece of advice that most people think of as weird because they feel like, "I want to just get moving, I want to get moving on my new thing, I want to throw out the old thing and get a new thing." But that's putting the cart before the horse. You want to know first what works on your old design, and what doesn't work, and why it doesn't work. Those are all very important questions to get answered, because otherwise what happens is that you may actually lose some of the features that worked well in the old design. And of the ones that didn't work, you know, maybe you'll have something different, but who's to say that different is better? So it's very, very important to do a study of the existing design. The existing design is your best prototype of your new design because it's already working, it has all the features, and it has all the users right there.

That seems common sense enough.

Sad to say, people often miss that.

In the past, you've suggested that 10 percent of the budget for any Web or intranet project go toward usability. Still a good rule of thumb?

That is still a good figure. Ultimately it should be higher than that, but in today's world, that's a good recommendation. The way to think of that is, really, that you spend 10 percent of your budget making sure you're doing the right thing, and then 90 percent on doing that thing.

Resources

Books

1. Designing Web Usability By Jakob Nielsen New Riders, 1999
2. Don't Make Me Think! A Common Sense Approach to Web Usability By Steve Krug New Riders, 2000
3. Information Architecture: Blueprints for the Web By Christina Wodtke New Riders, 2002
4. Web sites Nielsen's "Alertbox" www.useit.com/alertbox
5. Adaptive Path's incisive essays on information design, architecture and usability www.adaptivepath.com/ publications/essays/archives/index.php

Copyright (c) 2004 Ziff Davis Media Inc. All Rights Reserved.