Smalltalk Tidbits, Industry Rants

This is a wild story.

Law.com reports that software company Citrix is suing the Florida Department of Revenue to avoid paying taxes. Citrix's theory? "That all companies acting as purveyors of information and facilitators of communications, including newspapers, TV stations, telephone companies and computer companies, are constitutionally exempt from paying taxes of any kind," because "taxes have a chilling effect on the First Amendment right to free speech under the U.S. Constitution."

hmmm. double hmmm. triple hmmm, even...

John Brant posted this on cls today - IMHO, it demonstrates one of the truly cool things about the RB:

One thing that the RB has that I haven't seen in any Java refactoring tools is the generic parse tree rewriter (I've never programmed in Java, so I might be missing something obvious). Using the rewriter users can define their own transformations. There was an experience report at OOPSLA '02 where Will Loew-Blosser used the rewriter to change the data access layer: The report. Using the rewrite rules he was able to make the 16,000 changes 34x faster than he would have by hand.

Now that's productivity

That's the only way I can describe this web revival of old BBS stuff. Follow this link to get the story:

Welcome home! The new ANSI / BBS stylesheet is now live. MarkupLanguage.net, the latest of Mozquito's Webaccess developments, will be used for the revival of the StadtNet community system and is planned to initially run the forum, as many BBS systems had news forums at their core. It will be linked off Stadtnet as the 3.0 version of the "SN Foren", after having reached agreement with Joerg Stengel, founder of StadtNet, to restart the operation of this community system based on Mozquito software.

Call Dr. Emmit Brown!

Although it's not exactly clear to me what Oracle expects. See the story. Excerpt:

According to this article: "The proposal outlines a standardized way to merge Java programming tools from several companies into an integrated development environment (IDE), which would then allow access to all the tools through a single interface. Oracle intends to present a draft of the specification by March next year through the JCP. Java backers such as BEA Systems, Borland and IBM will contribute to suggested add-ons and incorporate approved updates to the specification to ensure it will work with their products. Oracle's stated goal is to codify the mechanism for plugging together different Java programming applications. Once standardized, an application developer could use a single Java IDE and be sure that an application for testing Java code, for example, would work glitch-free with tools for program design and source-code control."

This sounds an awful lot like a description of Eclipse to me.

From the VW Wiki:

Originally StORE was known as Bernstein but it was extensively revamped for VisualWorks 5i's namespaces and parcels. Members of the team wanted to call it StASH for Smalltalk Archival Source Holder but certain nameless senior members of engineering with a vague memory of the seventies poo-poohed the idea.

Chuckle. Some of us even remember that the working name for class ApplicationModel was GluePuppy

I saw this story at slashdot, and my first thought was that it was a reference to the Starcraft game, not to UFO's and such.

I got this today from a customer. Now, I've dropped the company names and industry sector, but this is a perfect example of what I talked about here.

Our business and technology bigwigs went through a 3-month review of our business needs and the system itself to see whether a migration is necessary. Nearly everyone agreed that migration would waste a lot of money to achieve less than stand still. It is now accepted that both VW and GS/S are the cornerstones of the system's success, and that by mirating/rewriting we would lose the market share and ability to deliver. For our sector look no further than (our prime competitor) who had a competitive system to ours. About 3 years ago, a newly appointed CIO decided to do away with Smalltalk because he had never heard of it. It took them 2 years to unwind the system, whilst the unsuccessful rewrite in Java failed miserably. We now employ most of their Smalltalkers (who have a solid background in our business domain - a big asset) and their end user support personnel. The outcome? They have lost a BIG profit making business completely, which now in the times of reduced business could have put their entire organisation into profit.

That's exactly the risk I referred to in migrations. Sure, Smalltalk is a niche product when compared to Java or .NET. However, deciding to do a migration based solely on the niche status of a technology (any technology) without regard to the risk is just strange. If nothing else, there is a huge opportunity cost involved in having a team do a standstill migration - what business value could they be adding instead?

I decided that the list of tech blogs I follow wasn't really expansive enough, so I added links to the various other industry and Smalltalk sites I follow. The sites listed don't encapsulate all the things I follow (for instance, I specifically omitted politics), but it's a pretty complete list of the places I get ideas and content from. Enjoy.

That interest in the Cincom Smalltalk NC is high? We are getting 800 - 1300 downloads a month - I've been personally tracking the numbers since September. Those are very good numbers, IMHO, and represent a good level of interest in Smalltalk.

No sooner do I re-enter the dynamic/static fray here when this shows up in comp.lang.smalltalk:

As we all have found out (latest when reading this thread ;-) is that there are at least two big philosophies in programming languages: static typing and dynamic typing. While many people make a white/black distinction between these two claiming one or the other side is inherently bad, I cannot agree. IMO programming languages are all in an early stage of development, and I feel disadvantages in all of them. ATTENTION static typing stuff ahead: Regarding generics, while uselesss for dynamic type languages, are important in statically typed languages, as they are the only way to solve certain problems. E.g. you have an AbstractBuffer superclass with ByteBuffer and IntBuffer subclasses. You will then can write a

int getNext(); in IntBuffer and a byte getNext();

each dealing with its particular buffer. With generics, there is only one method. I agree fully that Generics do not solve everything. Another big issue is that types cannot be refined in subclasses (this is even more painfull than the problems solved by generics). A good example is Object.clone(); Any subclass can override, but they all are forced to return Object instead of their type, leading to ugly code like e.g. ByteBuffer copy = (ByteBuffer) myByteBuffer.clone(); //yucc This one is particularly hurting when designing class hierachies. Also, it apparently drops back to runtime type checking, static types are merely a burden. (BTW: these two problems are on (1) and (3) in the java RequestForEnhancement list). Apparently all this is absent in dynamic type languages.

On my conference call. There were some positive things (and interesting things) that came out of it. The analysts we spoke to did say the following, quoting from some of their older papers (1997 and 1999):

1. Smalltalk is a niche
2. niches are bad, because you get less vendors supporting you, higher risk of being orphaned, and harder to find and more expensive people
3. Don't stay in a niche unless it offers you compelling technical advantages that outweigh these considerations
smalltalk has strong advantages, most notably in the area of complex and changing requirements And to quote one of the other people on the call from Cincom (i.e., one of the more level headed ones who did not lose his temper): I'll take that. I'd take it as an advertising slogan: If your requirements are simple and never change, then you don't need our products.

Which is the summary of what we were told about Smalltalk. We were also told that they (the analysts on the call) had not looked at VisualWorks in over five years as well. Draw your own conclusions from that next time you see advice being proffered from that direction.

Dynamic typing is still able to raise controversy - check out the archives of the USENET comp.lang.smalltalk newsgroup and you'll see what I mean - threads can pop up and linger for weeks (even months!). So it's always useful when someone discusses the issue intelligently. Have a look here - tip of the hat to Why Smalltalk for the link.

Anatol Fomenko wrote: ...Why Smalltalk is dynamically typed, and why it does not affect negatively the stability of the large Smalltalk applications?

Rephrasing slightly gives the two following questions:

1. Why is Smalltalk dynamically typed?
2. Why does dynamic typing (as done with Smalltalk) not negatively affect the stability of large applications?

The first question would be better answered by Alan Kay and the Smalltalk team more than anyone else relaying their reasons. Alan Kay has described some of his reasoning in the book "History of Programming Languages II", various OOPSLA & Smalltalk talks, and other sources.

In my own words, the main reason is that Smalltalk has a simple and powerful concept of building software out of Objects that send Messages to other Objects. This is more powerful in both the small and the large than a language that adds the additional concept (and constraints) of "Types". Note that biology has Objects (e.g. Cells) but no Types, Humans are Objects but there is no compile-time type-checking between humans, and even electronics do not have compile-time Types: you can plug the GROUND pin into +5V if you want although some physical constructs will discourage (but not prevent) it. These highly scalable areas (life, civilization, electronics) have done fantastically without the support of Types, growing in orders of magnitude of functionality. Alan Kay (who has a biology background) leveraged insights on how cells and other highly-scalable areas could scale, and applied them to Smalltalk. The main concept was membranes/encapsulation and little else was needed.

The second question may partially support the first. The short answer is dynamic typing can scale well because one tends to create much less code as the system grows bigger. As the system grows, objects will get reused in many different situations for which they work well, and the layers of "membranes" allow clients to not worry about internal details very much. It still requires a good architecture to build a big system, but the generalizable functionality of the objects is helping a lot with the design/implementation.

There's not a lot to disagree with there, IMHO. Go read the whole thing

That I laid out here. I got a comment that said, more or less: "1000 man years? No one does projects like that anymore!" Maybe they don't. However, there are plenty of projects like that dating from the early and mid (even late) 90's - a number of our Cincom Smalltalk customers have such large projects. How do I reach a number like that? Well, I have personal experience with one huge effort at a large (Fortune 100) firm. They have had 150-300 people developing the applications in question since late 1993 or early 1994. That's well over a thousand man years in at this point. It's hardly the only example of such a system either - I have friends on many other similar systems, and not a few of them have tried migrations to Java (watch for it - some will now try to go to .NET). Most of the efforts I know of failed; the ones that succeeded merely re-delivered the existing functionality in the (now trendy) language. The question remains - Where's the business value?

Scripting languages have more options than bisexuals. It's a reasonable, interesting article with one heck of a hook line to drag you in.

Update - Permanent Link

Here's a scenario where a company takes the all to common expert advice on technology. The advice I'm talking about comes from an analyst report dated September 27th - Leading Programming Languages for IT Portfolio Planning - but the advice is certainly not to this report of the authors of this report:

We recommend that enterprises consider a migration to Microsoft or Java languages and technologies. A qualifying factor in their assessment would be the degree to which their vendors align their strategies with Microsoft or Java

I've read the entire report; the reasoning used to justify this boils down to "Everyone else is using Microsoft or Java - you should as well!". I seem to recall a thing or two my mother said about "things everyone else does" - I'm sure you do as well. In a serious vein, what they don't talk about are the truly critical things you should look at in any large scale effort like the ones they advise:

1. What is the business value of the migration?
2. What are the risks inherent in the migration?

Those are the interesting issues, and the report ignores them. Perhaps that's due to the numbers that they themselves have turned up. Let's look at the two main issues by going through an example. I'll posit an IT shop with a large application, into which 1000 man years of effort has been spent to date. Now, let's posit that the developers will be equally productive in (insert popular technology here) as they are in (insert non-favored technology here). This is granting a lot, but it's very illustrative of the problem.

Assumptions

Time to rewrite application	1000 man years
Cost per man year (US)	$140,000 (fully loaded)
Cost per man year (outsourced overseas)	$50,000 (fully loaded)

Totals

US based cost	$140,000,000
Outsourced cost	$50,000,000

Now, that's a lot of money. And bear in mind, at the end of the exercise, the IT shop has the same application they started with, with the same features - i.e., they spent all that money to stand still. If the customer used consultants for this task, the costs could go even higher - the hourly charges (using a mainstream consulting firm) could easily top $100 per hour. But at the rate of $100 per hour (2000 hours per year), the cost using consultants could easily reach or exceed $200 million USD.

Can someone explain the business value of following this advice? I'm not seeing it. The costs are enormous even if we assume a 100% success rate in such migrations, and the customer will end up with the same functionality they started with.

But it gets better. If you look at the success and failure rates - from these media reports: 40% of all IT projects fail, but that 70% of Java projects fail. And according to that article, the expectation is that .NET projects will fail at the same rate. So let's go back to the example:

1. The recommendation is that you migrate a large (successful) application
2. The cost of migrating will be at least $50,000,000.
3. In attempting this migration, you are only 30% likely to succeed

That's right - the expert advice is suggesting that you spend truly enormous amounts of cash in an effort that is only 30% likely to succeed. I'll ask the question again - where's the business value?

The interesting thing is that I'm not the only one who has expressed this opinion on rewrites. Joel on Software has done so here and here. There are at least two very large, very public examples of companies following the advice laid out in the report:

1. Netscape, with the Javagator project
2. Corel, with the WordPerfect to Java project

All the things I said above applied - and both projects crashed and burned. Netscape actually didn't learn, and rewrote from scratch in C and C++ - and their market share went from 50% to 3% during the four years that they released no new versions. Corel nearly died in their attempt, and was severly weakened.

I've seen this in my customer base as well - many attempts to migrate existing, working Smalltalk applications to Java. Some succeeded, most failed. Even the successful ones spent huge sums of cash, and ended up with exactly what they started with. The opportunity cost of that is huge - what new value could their IT shops have been providing had they not been on a project to duplicate existing functionality?

That's quite a ramble, but my question is the same - What is the business value derived from following the migration advice?

Here's an interesting piece over at Gordon Mohr's blog on what I can only call cybernetic research:

Two great tastes that go great together: rat brains and computer circuitry. Scientists at SUNY-Brooklyn have wired a computer into a live rat's brain, allowing the researchers to remotely "steer" the rat, while a researcher at Georgia Tech has grafted rat neurons into a robot, allowing the rat brain remnants to steer the robot.

Quick, somebody go check out Cyberdyne systems! Just in time for next summer's T3

I read about this a few weeks ago, and got reminded by David Buck in email: There's a slimy little outfit called PanIP that is doing the following:

PanIP based in San Diego is suing 50 small businesses (so far) over patent infringements. They claim to have two patents - one on "using graphical and textual information on a video screen for purposes of making a sale" and one for "accepting information to conduct automatic financial transactions via a telephone line & video screen". Virtually every e-commerce site would be infringing on these patents. Rather than take on the big players, this company is suing small businesses knowing that they don't have the financial means to defend themselves in a patent lawsuit and would rather pay the $5000 licensing fee to get PanIP out of their hair. A coalition of companies is fighting back to take this to court and have the patents revoked. This is pure insanity. Related Links:
http://www.youmaybenext.com/
http://www.geek.com/news/geeknews/2002may/gee20020517011773.htm
http://yro.slashdot.org/yro/02/11/05/2121249.shtml?tid=155

And here's the url for the ethically challenged folks themselves: http://www.panip.com/ Even unto the age of the internet, PT Barnum's spirit lives.

I saw the film last night (the midnight showing) and again today with the wife at noon. It was well worth seeing twice. There were a lot of very funny lines for Gimli - and the action/combat sequences were very, very well done. One of the things I had wondered about was Gollum - how well would the CGI work (after the disaster that was Jar Jar Binks... ). Not to worry. Andy Sarkis (the voicing) was marvelous, and the movements and emotions were very well done - he seemed very lifelike. I also really, really liked what they did with Theoden - his transformation from the doddering old man under Grima andn Saruman's spell back to the commanding King. There were more liberties taken in this film than there were in the first - but again, I thought they mostly worked out. I was unsure about one of them (involving Faramir) after the first viewing, but I was ok with it after the second. highly, highly recommended

Why Smalltalk is helping out in my efforts vis-a-vis the analysts who cover application development issues. Let's hope that they look at their own stats a few more times - maybe enough to actually take them into account - and reconsider their advice that companies migrate off (insert anything but .NET and Java here) to Java or .NET.

Have a look at the Why Smalltalk page on proposed "Powered By Smalltalk" logos. There are some pretty nifty ones there - and a pretty fair bit of effort has gone into it. Tip of the hat to Jason Jones of Why Smalltalk for highlighting this.

Looks like Sun has pulled the plug on the Dot-Com Builder website. I remember seeing the ads in airports - the "We put the dot in dot-com" line. I especially remember the one showing a little girl's face, tongue in cheek (that bulge serving as the dot). They all disappeared around the time of the bubble burst; now the website is going. Tip to the Register

Sun Microsystems is 'retiring' its Dot-Com Builder website. Killing it, more like: the site shuts on December 23 and content is to be divvied up between other Sun developer web sites. Links to the old site are maintained through the Wayback Machine.

I saw this item on the Fuzzy Blog this morning, and had to laugh - that's pretty much my office garb when I'm not traveling, and yes, people stare - in my case, when I walk my daughter to the bus stop.

This will probably put a chuckle on people's faces this moring. When you work from home there is always the question of "What do I wear today?". I mean I don't think anyone ever wears exactly what they did in the office. Not that I'm saying that you want to slack off and be a slob or anything (and, no, I'm not a slob who is trying to cover up for himself) but it's always a bit different. Personally I tend to my "summer hacker" outfit year round i.e. Bermuda Shorts and a shirt.

It's kind of a relief knowing that I'm not the only one doing this.

to the midnight showing of Two Towers. I have a small quibble with one of the liberties they took, but that's ok - the movie was fast paced, and grabbed hard. Well worth seeing - and I will be again tomorrow, with my wife and her office...

And with better information than I had! Over at Gordon Weakliem's log web log are some thoughts from another veteran of STIN. I hope he didn't have to pull the 12 hour days us Booz-Allen guys (my employer at the time) did. I was wrong about Hugh (famous for HuBol) as well - I guess we all just thought he was near retirement.

Oh, yeah, TravelBase! I used to work in the basement of the STIN building with Hugh Wilkie, the creator of "HUBOL". I've heard it described as a blend of FORTRAN and COBOL with bit shift and rotate operators thrown in. Never saw any actual HUBOL code, but it was still kicking around at the time.

TravelBase was really my first death march experience. Always nice to know someone else remembers it...

Only 6 1/2 hours to LoTR - The Two Towers

I have a conference call with Feiman and Driver (Gartner) on Thursday this week. If anyone has suggestions for me, I'd love to hear them

David Brin writes a very interesting critique of the entire Tolkien mythos. I was tempted to lift a quote from his third page, but no - read the whole thing yourself - it's well worth a few minutes reflection.

Web Services according to Jon Oltsik. And this time, Smalltalk is prepared - VW 7 shipped with full Web Services support. From the article:

This was a tough call for me. I truly believe Web services will improve application development and integration and has the potential to change the way companies use software. Even so, the technology still remains very immature. XML (Extensible Markup Language) is a good foundation for standards, but data definitions and business semantics remain a battleground. Web services also suffer from a lack of application-level security

Last July, I had kind of an epic brain cramp with one of the internal wikis - see the post I referenced for details. The cool thing is, the server has been running nicely since then - and I've had to add a new wiki to it - without any hiccups. That's 5+ months of uptime, a pretty good track record.

Why Smalltalk has the goods on the upcoming ESUG 2003 Conference. I would love to go this, but the last week of August is just a bad time for me - that's when my daughter starts school, and it's time to meet her new teacher. This looks to be packed with good stuff:

Continuing a strong tradition in organizing Smalltalk events for Smalltalk practitioners, this year promises to be very exiting. ESUG organizes a Smalltalk Joint Event comprised of Camp Smalltalk, Advanced Seminars and a Research track:

1. An official (and free) Camp Smalltalk, starting the week-end before and continuing the rest of the week.
2. The Advanced Seminars, where invited speakers and technical presenters show concrete applications and research using Smalltalk.
3. ESUG Research Track: an academic research track with top-of-the-line program committee. A separate call for paper was issued for this event, which is attached to this mail at the bottom.

The event will be in Bled, Slovenia between August 23-29. Follow the link to Why Smalltalk for details and contact information!

Application Analysts - How Reasonable? Update - Permanent Link here

In the IT industry, the various analysts have a large amount of influence. They have analysts who follow all aspects of the IT industry - server platforms, development tools, operating systems, the whole gamut. This raises a question though - with how influential they are, how accurate are they? It should be pointed out that I am hardly an unbiased observer - as the Product Manager for Cincom Smalltalk, I am obviously going to have a distinct point of view on what people should use for developing software. Still, over the last few years I've watched the analysts go negative on Smalltalk, and I decided to start looking at what they say and how they justify it.

The best place to start is with an analyst report put out at the end of September: Leading Languages for IT Portfolio Planning. This report looked at what development languages are in use, and what development languages are likely to be in use. Additionally, it makes recommendations on what languages should be used - and also on which should not be used. The report was written by the analysts at the analyst firm in question who follow development tools. Here's the table of language penetration they came up with in their survey:

Language	IT Shop Penetration
Java	70%
VB	70%
Cobol	45%
C++	45%
PowerBuilder	20%
Delphi	10%

These are the languages that they classify as mainstream. They consider Java, VB, and (to a lesser extent, C++) to be mainstream, with C# soon to enter that state. To quote them, they say the following about this table:

"Other languages and integrated development environments such as Pascal, Delphi, PowerBuilder, Smalltalk and Natural have already been pushed into a niche area, where we expect them to remain. We recommend that enterprises consider a migration to Microsoft or Java languages and technologies. Qualifying factor in their assessment would be the degree to which their vendors align their strategies with Microsoft or Java. Bottom Line: Enterprises must balance specific language requirements with the mainstream critical mass of skills. When possible, mainstream IT enterprises should first look to align application development initiatives with Microsoft.NET or Java-related technologies, tools and programming languages. They should consider niche tools only when the risk is far outweighed by the limited use of these specialized toolsets"

What they recommend here is that IT shops should strive to be in the mainstream, where it will be easier to find skilled developers and tools. That sounds reasonable, until you notice these comments made to the press by an analyst who follows aplication development tools: made in an interview:

"A September survey by the Gartner Group, a technology consulting company, revealed that approximately 40% of information technology (IT) projects do not produce their intended results, an indication of how badly some companies need the consulting services of a tech-savvy CPA. Gartner's prediction that global IT spending will reach $3.3 trillion by 2002 highlighted the broader implications of this finding"

This is in line with other reporting; the statistics on project failure tend to range between 30% and 40% depending on whom you ask. Where it gets interesting is when you see the same analyst quoted above saying this to the press:

To date, around 70 percent of initial Java implementations have been unsuccessful, according to new research from Gartner Group.

"An inordinately large number of large-scale Java projects have been failures," said Mark Driver, Gartner research director for Internet and ebusiness technologies.

However, Microsoft shouldn't draw any comfort from those figures as it seeks to promote its .NET technology strategy either. In all likelihood, the failure rate for early implementations of .NET systems will be similar, Driver said.

So let's recap here - analysts recommend that you use a mainstream language for your development - Java, VB, or a .NET language is what they mostly recommend. They particularly recommend that you migrate projects from Pascal, Delphi, Smalltalk (etc.) to Java or .NET post haste - and they do so even though their own research shows that the failure rates for that decision are inordinately high.

It's at this point that I have to question the value of their reporting. They have the failure rates in front of them - they are quoted reporting them. And yet, in an official publication, they recommend that any shop using a non-mainstream tool should migrate in order to mitigate risk. I have to ask, what risk? The statistics show a 40% failure rate overall for IT projects, but a 70% failure rate for Java (and likely .NET as well) projects. That means, logically, that starting a project in anything but Java or C# has a lower than 40% chance of failure, while starting in Java or C# yields a 70% failure rate. This isn't my Smalltalk advocacy speaking; this is their very own data speaking. In full view of all this data, these statements have been made to my customers:

"Smalltalk has become a niche technology that will be used by less than five percent of enterprises for another few years, mostly in the maintenance mode."

"There are one or two reputable vendors that still support Smalltalk (e.g. IBM) but they do not and will not introduce significant enhancements, recognizing the niche status of Smalltalk."

"Smalltalk applications should be relegated in the maintenance mode: do enhancements, do not do major new development!"

"As a strategic, long-term direction enterprises should choose a mainstream technology: either (or both) Java or/and Microsoft technology."

Now, these comments were made at a point when Cincom was about to release VisualWorks 7, IBM was about to release VAST 6, Gemstone was about to release Gemstone/S 6, and Object-Arts was about to release Dolphin 5 - all with major new features you can easily read about by visiting their sites.

So there you have it - the analysts recommend that you switch from technologies that have (at least) a 60% rate of success in project usage, to technologies that have a 30% success rate in project usage. Draw your own conclusions

I picked this link up from David Buck in the comments to this post. It's worth it's own extraction, IMHO.

Java's marketing issues aren't an insignificant problem, said Borland's Shelton. In a recent survey Borland conducted of "C-level executives," i.e., top corporate officers, the predominant response from those surveyed was "Java is over." Respondents perceived that Java had been pitched to them as a panacea for everything, and once it failed to deliver on that promise, they became disenchanted with the platform and receptive to the promises of Microsoft's .Net. "How have we, as an industry, led our C-level people to believe Java was going to solve the common cold and cancer, to the point where they're now so fed up with it? That's the problem we need to solve," Shelton said.

The amusing part to me is, here's a bunch of people who seem to recognize that Java was not a silver bullet. And yet, there they are ready to treat .NET as one. The industry continues to not get it.

Go read this now. I was rolling on the floor - I have the same problem ordering extra sauce!

There's some question as to whether IE usage is as high as reported - well over 90% according to this story. The question is, how many of the reported IE users are in Opera, Mozilla (et. al.) and simply reporting as IE to get around sites that only want to support IE (an astonishng number of sites are like that)?.

OneStat.com said that since it last published data in September 2002, IE 6 has picked up an additional 5.3 percent, moving from 52.3 percent to 57.6 percent share. IE 5.x has 35.2 percent of the market according to OneStat.com's metrics, while IE 4.0 holds 0.9 percent. The firm also showed Netscape 7 picking up 0.1 percent from 0.5 percent to 0.6 percent, with the Netscape offering overall holding 3 percent of the market -- making it second place in the browser wars. Meanwhile, the firm said its data shows Mozilla holds global usage share of 1.1 percent and Opera 6.0 holds onto 0.8 percent. It's in these details that OneStat.com's data begins to vary slightly with W3Schools. As of October 2002, W3Schools said Internet Explorer 6 holds 45 percent of the market, IE 5.x holds 46 percent of the market, and IE 4.x has 2 percent of the market. The methodologies of both firms are similar. OneStat.com said a global usage share percentage for a particular browser is generated by measuring the percentage of Internet users -- and which browsers they use -- that arrive at sites using one of OneStat.com's services. W3Schools does the same with data generated by TheCounter.com, a service run by internetnews.com parent Jupitermedia. However, Unix Systems Administrator Ben Rosenberg said such statistics should be taken with a grain of salt. Browsers use an identification string to identify themselves to Web sites, which is how counters and metrics firms can generate data about them. But a browser like Opera defaults to IE as its user string identification if the site is not configured to identify Opera. And both Konqueror and Mozilla users can change their user-agent string to IE to make sites accessible that are only geared toward IE. In fact, there is an add-on for Mozilla and Netscape that allows users to change the user string identification on the fly.

I doubt it makes a huge dent in the numbers - IE is certainly the market leader. But it does point out the limitations behind this type of research as the client side tools get more and more configurable.

The analysts recommend migration off of languages that don't appear in their top five (except for C# and VB.NET). Wait, call out the media - they just stated that PHP and Perl are doomed as well. And Python. And Ruby. Heck, C as well. Shhhhh. Better not tell any of the open source developers. What complete idiots.

I just got a copy of this Gartner report: Leading Languages for IT Portfolio Planning
Here's the money quote at the end:

Other languages and integrated development environments such as Pascal, Delphi, PowerBuilder, Smalltalk and Natural have already been pushed into a niche area, where we expect them to remain. We recommend that enterprises consider a migration to Microsoft or Java languages and technologies. Qualifying factor in their assessment would be the degree to which their vendors align their strategies with Microsoft or Java. Bottom Line: Enterprises must balance specific language requirements with the mainstream critical mass of skills. When possible, mainstream IT enterprises should first look to align application development initiatives with either Microsoft.NET or Java-related technologies, tools and programming languages. They should consider niche tools only when the risk is far outweighed by the limited use of these specialized toolsets.

So, in their discussions with clients, they recommend that you get off Smalltalk (and a raft of other technologies by implication). They do so even in the face of their own research. So they know that the recommendations that they give will lead to - in their words an inordinate chance of failure. That's expert analysis? That's worth paying for? Exactly what value is being provided, given that following this advice will lead to a large increase in a shop's risk? I, for one, am tired of this 'expert' advice.

The new Netscape - and apparently the current Mozilla as well - banish popups. Read the Register story.

Three More Days!

I posted yesterday on the wonderful logic behind the language/development analysis. Well comes today some more information: These are the language usage statistics that Gartner's analysts believe to be true:

Language Usage Statistics

Java	70%
VB	70%
Cobol	45%
C++	45%
PB	20%
Delphi	10%

This is supposed to represent language penetration. According to the report, if your language isn't there, you better stop using it. Looks like we have some news for all those Perl, Python, and C developers - time to migrate! But hey, if you are using Cobol or PowerBuilder, you are up to date and in the mainstream. Now recall, the analysts in question produced this report - which states that 70% of all Java projects fail. And in the comments to this old post, you get this lovely direct quote:

"Smalltalk has become a niche technology that will be used by less than five percent of enterprises for another few years, mostly in the maintenance mode." "There are one or two reputable vendors that still support Smalltalk (e.g. IBM) but they do not and will not introduce significant enhancements, recognizing the niche status of Smalltalk." "Smalltalk applications should be relegated in the maintenance mode: do enhancements, do not do major new development!" "As a strategic, long-term direction enterprises should choose a mainstream technology: either (or both) Java or/and Microsoft technology."

So I posted some fun with statistics yesterday. Let's revisit those numbers under a different set of assumptions. Recall that

1. Gartner's own numbers claim that 40% of all IT projects fail - and this is in line with other industry research - story here
2. Gartner's own numbers claim that 70% of all Java projects fail - story here
3. Gartner has been recommending (since 1997 at least) that IT shops migrate from Smalltalk to Java

In the post referenced above, I assumed 50% market share for Java. That's way, way too high - between existing projects, legacy work and what not, let's assume some different numbers: Assumption 1

Java has 10% market share in development

40% (failure rate) = 70% (Java failure rate) x 10% (Java share) + N x 90% (other dev languages share)

That boils down to N = 37% failure rate for all languages other than Java, assuming a 10% share

Assumption 2
Java has 20% market share in development 40% (failure rate) = 70% (Java failure rate) x 20% (Java share) + N x 80% (other dev languages share)
That boils down to N = 32.5% failure rate for all languages other than Java, assuming a 20% share Assumption 3
Java has a 30% market share in development 40% = 70% x 30% + N x 70% That boils down to N = 27% failure rate for all languages other than Java, assuming a 30% share

---

But you get the point - Gartner's analyst on this has been telling people to migrate from Smalltalk to Java for at least 5 years now, even though he has been in full possession of these facts. He has been telling people that Smalltalk vendors are not adding new features to their products - he said so earlier this year, in fact to some of my customers - even though he knew about the upcoming releases of VW 7, VAST 6, Gemstone 6, and Dolphin 5!. In short, the Gartner analyst who follows Smalltalk doesn't do minimal fact checking - and is in a position that enables him to do huge amounts of harm. Now, I don't begrudge an analyst expressing the opinion that Java skills are easier to find, or that Smalltalk is more of a niche tool - what I mind is the blatant lack of fact checking - i.e., the assertions that no new work is being done by vendors. I mind the fact that he tells people to migrate off, knowing full well that this increases the risk of project failure dramatically. In short, I'm sick and tired of experts deciding that Smalltalk is dead, and then making sure that it's true by pounding us in the head with a shovel.

has gotten some attention. This is the original post from Joel; it seems others are weighing in on the topic. For instance, I picked this up from Gordon WeakLiem, who got it from comments elsewhere:

Personally I think the heyday of really hard programming were the early 1960s, when the first big distributed electronic commerce systems like SABRE were being built. American airlines contributed the 1960 equivalent of a billion dollars to build SABRE over 5 years (I don?t know what other airlines spent). I believe these numbers and timeframe are similar to what Microsoft spent to develop the entire .NET platform. It?s hard to believe that if the airlines wanted to build a reservation system today they couldn't use mostly off-the-shelf software and get it done for 1/10 that cost (or less, I?m an optimist). Even if it feels like you have to run to stand still in this industry I think it's easier than ever to get things done.

Gordon then goes on with this:

Talking about the travel industry is always a good way to get me to pay attention ;-) If you want a good read on the construction of SABRE, chapter 2 of the book Hard Landing has the story. In fact, Sabre was the 2nd system American Airlines tried, another company (Burroughs, IIRC) tried and failed before IBM succeeded. I still work in the industry, and the amount of effort involved in making changes on the mainframe is absolutely incredible. One of my co-workers who's worked on the mainframe tells me that he used to maintain code where the last modification date was before he was born. Now that's legacy code.

I spent some of my own time in a Sabre offshoot - a wild attempt to rewrite the TravelBase system. It had started out, as all of the Sabre system did, as a mainframe system. There were stories galore there; it was written in HuBol - named after Hugh, who's office was in the basement. If you had a question, you could ask Hugh - if he felt like answering, he would. Being close to retirement, his concerns didn't always match ours.... This was an attempt to rewrite the system - out of the mainframe and the Cobol offshoot, and over to OS/2 and C. You haven't seen hell until you've seen a bunch of Cobol programmers trying to do their first even driven, GUI based system in C. I spent 6 of the longest months of my life on that death march, and all I have to show for it is the wrist wrest under my keyboard, which has the official launch date - June 2, 1992. Don't you believe that the system actually worked then....

I posted earlier on the Gartner reporting - they have statistics showing an overall industry failure rate if 40%, and of 70% if using Java. I posted this to comp.lang.smalltalk, and John Brant applied the following:

40% of all projects fail and 70% of all Java projects fail, so what's the
rate for non-Java projects :-)

40% * (# of projects) =
    (70% * (Java market share) * (# of projects)
        + (non-Java failure percent) * (1-Java market share) * (# of
projects))

If we assume that Java's market share is 50%, then we get:

40% * (# of projects) =
    (70% * 50% * (# of projects) + (non-Java failure percent) * 50% * (# of
projects))

This can be reduced to:
    40% = 70% * 50% + (non-Java failure percent) * 50%
==>
    40% = 35% + (non-Java failure percent) * 50%
==>
    5% / 50% = non-Java failure percent
==>
    10% = non-Java failure percent (assuming 50% market share)

So, if they actually believe their own statistics, they

1. Recommend that you use Java instead of (say, Smalltalk)
2. Based on their own stats, they know that following their recommendations has a 70% chance of failure
3. Based on their own stats, they know that not following their advice has only a 10% chance of failure

kind of sums it up, doesn't it? They don't check facts - they make claims that Smaltalk vendors aren't upgrading their products, when this is demonstrably not the case. They tout statistics that they can't possibly believe - if they did, they wouldn't make the recommendations they make. None of this is hard to check - heck, I did nothing but follow links on the net to gather this information. Seems to me that it's time someone pointed out that the emperor has no clothes

InfoWorld interviews the intrepid Scott McNealy. He goes on reasonably enough for awhile, and then gets here:

IDGNS: Do you think you waited too long to bring out a Linux on Intel server? Did you wait too long on an Intel-based system because SPARC is in your blood? SM: If you look at what Sun has done that no one else has done in the Microsoft era, we have developed a real live developer community. There is the .Net developer community and the Sun ONE (Open Net Environment), Java, XML developer community. After that, there really aren't many interesting developer communities. The rest are small and shrinking. We are doing everything we can to provide as much opportunity for the Sun ONE, Java, XML community. We think we can leverage the 32-bit price-performance points of the low end of the Intel space to make Sun ONE a better price-performance environment for some areas

Apparently, McNealy is unaware of all the Perl and Python work going on. Purposely unaware, maybe. Clueless, more likely.

If not, why not? The new STIC has been put back together through the efforts of Allan Davis and Jason Jones (who also sponsor the excellent Why Smalltalk site. Individual and corporate memberships may be purchased online - and individual members will get a nifty CD with all of the free and non-commercial Smalltalks on it. Join now!

One of my intrepid readers linked to my earlier Rational post, and, there are comments on this weblog.

I'd give rational more credit: a) They may have used eclipse because they recognized a need to have a better technical foundation for their product; b) they wanted to be more appealing to IBM; c) they recognized that they need to become more agile and saw eclipse as a step in that direction (it is); d) they relized that there was a market for a common tool set for dot net and j2ee deveopment in large organizations. e) a+b+c+d. I'm amazed that there was not more Slashdot interest in the Togethersoft connection. As half-troll pointed out, Borland just agreed to buy Togethersoft. Togethersoft Control Center offer support for multi-language, model driven development -- many Java developers prefer it to the WebGain and/or Rational tool set. Of course, Together Control Center addresses problems not normally encountered by the slashdoters such as deploying to multiple J2EE servers, object modeling, complex documentation that is generated from code, etc.

Read the whole post, he's got interesting things to say.

Projects don't fail for technical reasons. They fail for people/communication reasons What the analysts and pundits don't take into account is how technical choices aid or impede communication. One of the largest problems faced by IT shops is churn - there is always a new silver bullet technology that will solve all the problems. What this means is that most IT shops never gain enough expertise in any given set of tools to be truly productive; they are always moving on to new tools that will theoretically solve the problems better than the old ones. What developers need is tools that will help improve communication - both between themselves, and between their customers. If the technologies they choose to implement in are complex, then developers will consistently be speaking jargon - solving technical problems instead of business problems. If, on the other hand, developers are given tools that are not complex, they will spend more time solving (and discussing) business problems instead of technical issues. To use a trite example, simpler tools will allow developers to discuss requirements instead of compiler errors. Tools that mostly get out of the way and allow developers to express the intent of their customers are better; tools that layer on complexity and arcane rules stand in the way, and prevent or delay these expressions of intent. One thing to bear in mind is research from psychology. People can, on average, keep 7 (+/- 2) things in mind at once. Adding to the number of things that need to be kept in mind is not really possible - it's a human limitation. Choosing tools and/or processes that insist on throwing complexity directly in the face of a developer are as sure a recipe for failure as is possible - simply based on the reality of human capabilities. What does this mean? It means that choosing a strict process with many steps and lots of documentary steps is not going to work - the sheer complexity of the process will over burden the average developer. Likewise, choosing a set of development tools that require a steep learning curve will add problems. It will take longer for developers to internalize the rules well enough to be productive. It's a simple fact that projects will not normally be staffed with as many experienced people as you would like - so choosing technologies that make things simpler will get you started sooner. Where does this drive us? Processes - pick an agile methodology. A heavyweight, process driven methodology will wear developers down and prevent progress. Tools - pick an agile toolset. A set of complex tools will likewise wear developers down and prevent progress. What satisfies these? Visit the Agile Alliance for tips on process. For tools, take a look at Smalltalk. It's no secret that the Agile Alliance leaders are virtually all from the Smalltalk world; there's a reason for that. Smalltalk is simple and powerful. A beginner can learn it in minutes, and an expert can express a design clearly and tersely. One complaint you'll hear is that Smalltalk has a large learning curve. And Java or C# don't? Smalltalk has 5 reserved words and syntax that fits on a 3x5 card. C# and Java have 50+ reserved words and very complex syntactical rules. It simply is not the case that Smalltalk is harder to learn. If you want to beat project failure, the best way to do so is to use an agile methodology with Smalltalk.

Just follow this link to see what I mean. Scary, kind of...

Awhile back I had this to say about suggestions that Java, being mainstream, is safer than Smalltalk. Well, there's some fascinating numbers you can pull out on that. This Gartner information states that fully 40% of all IT projects fail:

A September survey by the Gartner Group, a technology consulting company, revealed that approximately 40% of information technology (IT) projects do not produce their intended results, an indication of how badly some companies need the consulting services of a tech-savvy CPA. Gartner's prediction that global IT spending will reach $3.3 trillion by 2002 highlighted the broader implications of this finding.

Now combine that with this Gartner study, in which they state that, to date, Java projects have a 70% failure rate - and that they expect the same failure rate for .NET. So, by using Java instead of what you already know, you increase your chances of failure dramatically. To see the weirdness here, merely note that they

1. Recommend that you use Java or .NET
2. Note that they report much higher failure rates for those decisions

and then decide whether they are worth listening to, since they apparently don't look at their own data when making recommendations. Heck, even political polling firms are smarter than that.

In response to my complaint here, (which I also sent in a slightly more judiciously worded email), I did get a decent response:

Me: Is the Java team completely blind to events outside the Java Universe, or is it worse than that? Sun: None of the above: we neglected to spend enough time supervising the hype generators closely enough. We know better, and I apologize for the indiscretion.

so I'll acknowledge the kind response. Now, let's see if they update the site....