|
"The best way to predict the future is to invent it." — Alan Kay |
 Alan Knight |
|
|
"The best way to predict the future is to invent it." — Alan Kay |
Alan Knight |
Copyright 1999 by Alan Knight
This is my internal conference report on Smalltalk Solutions 99. Contents consist of my opinions or my twisted interpretation of facts and do not represent official positions of my employer. You may redistribute freely but please do not modify.
"Q: Where's main()? A: New England." - Ron Charron
Links to other reports you can check out
"If there's one place I know I don't want to die, it's in a hotel room in Newark, New Jersey" - Lothlorien Homet, SIGS Books
Flew in to beautiful Newark Airport, along with a whole bunch of other once and past Object People. Shared a very nice limousine from the airport with Martin, Vassili, Chulaka, and Anthony Lander. Anthony reads the program and discovers that he's part of a panel on persistence on Tuesday. I'm on the panel too, and have no idea what we're going to be talking about. Should be fun.
For those of you who haven't been to Smalltalk Solutions, it's held at the Marriott Marquis in midtown Manhattan. This is directly on Times Square, and there's a Broadway theatre in the bottom of the hotel. The hotel itself is hollow, with glass-walled elevators. From the top, you can't even see the floor of the lobby, which is on the eighth floor. It's very impressive, and phenomenally expensive. The conference facilities are various rooms and halls scattered around the fifth, sixth, and seventh floors. There are rotating lounges/bars on the 8th floor and at the top.
"I'm sorry I'm not wearing spandex" - Eric Clayberg
The conference begins with pre-conference tutorials. The hotel is large, and our presence is totally dwarfed by the massive aerobics instructors' conference taking place at the same time. Spandex is everywhere. I went to Eric Clayberg's talk on Advanced VisualAge programming which had a lot of very very cool stuff, messing with internals and a very high-level audience helping contribute. Good nerd fun. Samples include using blocks as receivers in callbacks, a very neat trick for simplifying conditional loading of subapps, the fact that VA timers (CwAppContext>>addTimeOut:...) are better than delays, and interacting with the clipboard. There are pointers to some good stuff, including an icon-in-DLL freeware viewer on TotallyObjects (http://www.totallyobjects.com/vafree.htm). Much of what he talked about were things he learned doing VA Assist. It's a very nice tool.
Slides and code examples are available at http://www.smalltalksystems.com.
"The community isn't nearly as afraid as it thinks it is" - Kent Beck
In the evening was a Kent Beck Birds of a Feather session, supposedly about his new book, "A Sorted Collection" grimace. (I bought a copy, had it autographed ("Grow some Hair") and then somebody stole it). Actually, he hardly said anything about the book, "I found it a really interesting exercise to go back over my own thought processes and see their development. I don't know if it's interesting for anyone else to read, though") and it turned into an audience participation session on Smalltalk viability.
Tidbits: There were ~150 paid attendees, down from I think ~225 last year, Kent said he assumed SIGS wouldn't hold it next year (he was wrong, read on). There was a lot of moaning about pressure from management, pressure from developers who want to be doing the latest thing, fears of irrelevance, lack of new projects, etc. etc. Kent took a poll on how many people here expected not to be doing Smalltalk a year from now, and a handful of people put up their hand. He said that was interesting, because he expected two-thirds, which inspired the lead quote.
One interesting attendee was a Java programmer who had gotten very interested in design patterns, found that "it was obvious all the really interesting work was coming out of the Smalltalk community", started using Smalltalk and is now a major bigot. Lots was said, and lots of people got lots of things off their chest, but nothing was really resolved, other than to try to organize a plan of action on the VisualWorks Wiki.
Kind of a downer, but in a way it loosened up the rest of the conference. If you figure this is the last hurrah, it may as well go out in a major blaze of fun and Rick Friedman's money.
"RMI just isn't scalable" - Jeff McAffer
IBM's Server Smalltalk is some very cool technology. They've kind of bundled it in with a bunch of other stuff (including XD, which is also very cool), but it's quite interesting on its own. It's a mix-and-match distribution framework, mixing transports (TCP/IP, MQ, SMTP, whatever) protocols (HTTP, RMI, IIOP, ST-to-ST, etc.) threading mechanisms (pooled, single-thread, logical process), and marshallers. Go read about it.
Interesting questions/comments: Jeff strongly recommends IIOP rather than RMI for talking to Java, even though there's no GC. Server Smalltalk isn't a full ORB because - they don't provide most of the more sophisticated services (basically just naming), they don't have tools, e.g. IDL parser or repository. His expectation is that if you're using CORBA you're doing it because you already have an ORB and you want to use its tools. If you're just going Smalltalk-Smalltalk, you don't need an ORB.
Jeff will be putting up a lot of server Smalltalk examples, but was still looking for a site as of the conference.
In going to this, missed Jim Haungs' "Bottom-Feeder's Guide to Smalltalk", which seemed like a very cool look at the nasty internals and meta-structure.
"I've decided that I don't like PowerPoint" - Kent Beck
Kent's a really good speaker, so I'm spending a lot of time here just quoting him. Some of it is semi-quoted, i.e. I'm just writing down what was in my notes but the exact words are probably wrong.
"Keynotes are a good place to tell a captive audience something they don't want to hear. Here's one thing: we lost. We all had this dream of Smalltalk in the mainstream as the alternative to C++, and it doesn't look like that's going to happen any time soon."
Having said that, Kent will now proceed to leave us hanging and go talk about extreme programming for half an hour, which I'm not going to talk about, since we just had Ron Jeffries in at the TOPICS-98 conference. See http://c2.com/cgi/wiki?ExtremeProgramming or http://www.armaties.com for more information on XP in general.
A few gems/nice points
...Back to Smalltalk, and how we're going to survive. We've lost in the sense of cracking the mainstream right now, but that doesn't mean we're dead. Lots of small technologies have niches, or remain dormant a long time. We can't take on Java head on. We can't be the 800-pound gorilla. What we need is some segment of the market where we can be the 100-pound gorilla in the small cage.
For example, if you're building software for the Palm Pilot: a hardware segment (see the pocket Smalltalk section, Kent really likes it). Or a particular vertical market. Bond trading, or real estate. Attacking the Internet market is probably a bad idea, even though we could do it well. Linux: segmentation by OS. Just some segment, because the message of "best for everything" doesn't work.
"Smalltalk is wonderful in that we have a common culture - a very rare thing. We care about people. This west coast, hot-tubby kind of culture. Java has come a ways towards that, but made some fundamental mistakes. I mean, source code in files. How quaint. How 70's."
Of course, out of everything Kent said, the one really quotable line is "we've lost", and this pissed off a lot of people who didn't hear the whole talk (and some who did). Here's Kent's explanation on comp.lang.smalltalk.
My Smalltalk Solutions 99 talk began by declaring that Smalltalk had lost and the community had better get over it. "Lost" in the sense that we wanted Smalltalk to be THE alternative programming language to C++ for all kinds of application development. This led to amazingly scattered marketing and no deep and sustained penentration of any definable market segments.
My proposal was (is) that we stop fighting the war we lost and we start fighting a new one- finding a market segment we can dominate, where folks will say, "Oh, you're building XXX software? You must be using Smalltalk. You're not? You must be crazy."
Solve for XXX.
I believe this is Business 101- you have to have a home base where you are not scratching and clawing for each and every sale. You can use the profit margins from that segment to expand to other segments.
Apparently my comments pissed some people off. First time that ever happened to me...
-- Kent
"Another Smalltalk lead? I didn't think there was any activity on this Smalltalk stuff, and it's all I'm doing lately" - Katherine Sayn-Wittgenstein, TOPLink Sales/Marketing
Booth traffic was dead for the first while, but then picked up quite a bit. Looked like everybody went for lunch after Kent's talk. At 1:00 we gave a TOPLink presentation. We had fairly modest expectations, figuring there would be maybe a half-dozen existing TOPLink customers curious about the new features in version 5.0. Instead there were 23 people, many of whom had obviously never used TOPLink, and they all stayed the full hour and asked a lot of questions. We ended up revisiting TOPLink 4.0 features and promising a 5.0 features review in the session the next day. Vassili was looking for the room and walked right by, since he didn't think there could possibly be that many people at our talk. On the second day, we had another 14 people. The VisualWorks 5i/Linux presentation came after us, and they filled the room to overflowing (I'd say 40+ people).
Overall, the exhibits hall was fairly small, with about 15 companies represented (Applied Reasoning, Cambridge University Press, Chubb, Cincom, Gemstone, IBM, Instantiations/Smalltalk Systems, ObjectIntelligence, Object People, ObjectShare, Precision Systems, SIGS, Silvermark, STIC, Synchrony, Unity).
Manage to find Jeff Odell, and we come up with a list of topics for the persistence panel.
"We need to get a bunch of technical people from customers to come down for a day and vomit on us" - Jim Robertson, ObjectShare (ed: I think he means in a technical sense)
Various announcements that don't fit into the chronological model
Q: Guess which industry-leading product which uses emulated widgets?
A: Microsoft Word. Apparently all the toolbar widgets are emulated, since native ones take too much startup time.
VisualWorks 5i shipped this week. It's the first really extensive upgrade from ObjectShare in a really long time. VisualWorks 3.0 was really a fix-up of VisualWorks 2.5. This starts introducing significant new features, including Namespaces, XML support, a cross-development feature called OpenTalk (reminiscent of the cancelled "Firewall" product, but based on DST), and their own version control system (although they're not trying to push out ENVY, just provide an alternative, especially for the lower end and the disconnected).
They've also split it into multiple product lines. There's the free non-commercial, the Personal (US$495) Enterprise, and OpenTalk. I don't know what the prices on the upper editions are.
This is a new product from IBM, running on top of VA Smalltalk 4.5 (i.e. you don't have to wait for version 5.0). It's similar to Applied Reasoning's "Classic Blend", although different in detail. You create UI's in Smalltalk, using a special palette of what are, effectively, Java widgets. Then you can run an applet in a browser using Java that communicates with the Smalltalk server for domain logic. It's about a 230K generic applet that can do presentation, basic data validation and editing. Requires Swing.
For more information, see
As ObjectShare divests themselves of the last of their Java encumbrances (see http://www.objectshare.com/pr/p4jteam2SEAGULL.htm) Gemstone are still torn between their Smalltalk market and their push for an IPO as Enterprise Java Wanna-Beans. On the one hand, their home page, which doesn't mention Smalltalk anywhere, proudly advertises some of their major customers, including Smalltalk ones.
On the other hand, they recently announced a free Linux version of GemStone, which is quite cool. It's limited to two users, so it's effectively an evaluation, but it's still a nice step. Information is available through their Smalltalk back door at http://smalltalk.gemstone.com.
There were too many interesting talks to mention all of them. Here are some I didn't get to that I thougth were particularly interesting
"Why do we hold this as a separate conference anyway. Wouldn't it make more sense as part of Object Expo New York?" - Lothlorien Homet, SIGS Books
The aerobics instructors are gone, but today there's a champagne-tasting event taking place right beside one of the conference rooms. I'm having trouble remembering why I'm in a solitary profession that involves a lot of late nights hunched over keyboards. I'm also having trouble remembering why I like conferences that have booth duty starting at 8:00am. I do that, then skip the early morning sessions to have breakfast and make some notes for the panel this afternoon.
"We are colonizing the planet with people who can think about systems because they can program in Smalltalk" - Mark Woodman, The Open University
As a company, we're pretty focused on the corporate development market. Big companies with lots of money and little object experience. Nothing wrong with that, they're the ones with the money, but it does make us tend to forget our humble origins in academia. We particularly tend to forget that obscure island off of Europe where we have a branch office. This talk was a reminder that in fact interesting things do happen in these places.
The Open University is a distance learning institution based in England but serving all of Europe. They're quite large. They take in about £220M/year, teach 170 undergraduate courses, have 1000 professors, teach 170,000 students/year and have graduated, since 1971, approximately 200,000 bachelor's degrees. The average age of their students is 37, so it's not the typical university demographic. They are also the largest publisher in the UK. One of my favourite lines was "That's just our internal student newsletter, circulation 180,000".
One of their courses is "M206: Computing, an Object-Oriented Approach". This took ~50 person-years of effort to develop, and has about 5000 students/year (it actually had ~8,000 last year). It's a web-centred multimedia course, with about 300 web pages, a 400 page text, on-line conferences, 30 LearningBooks, 3CDs and 12 TV programs. There are affiliated organizations, which will be teaching this course currently being set up in Singapore, Hong Kong, and the US. They have two other graduate courses using VisualWorks.
The programming environment is LearningWorks, which is a derivative of VisualWorks that has been hacked to the point of un-recognizability to be a learning environment. They can restrict the visibility so that students only see very small class libraries at a time, even when debugging, share work over the network, save state, all in a very simple teaching paradigm. Very cool.
"We have no regrets. Choosing Smalltalk was the best decision I ever made"
For more information.
"In VA/Java, it feels like I'm writing Smalltalk code, but at about one-third the speed. Which is not that bad." - Kent Beck
Finally. The panel. Moderated by Jeff Odell, who's currently part of a start-up building health-care applications in VisualAge. Featuring Me, Anthony Lander (ex-Object People, now with Silvermark), and Timo Salo (IBM, Object Extender guy). I think we actually came up with a pretty good theme for this. Everyone is bored to death of basic discussions of persistence, how to handle caching, etc. etc. Even if they aren't, I am. Since everyone on the panel is or has been involved with a persistence framework, let's take that as a starting point. Assume you have a reasonably capable persistence framework, one that reads and writes objects properly. Now what are the issues you need to face?
There was some good discussion, but since I was up at the front I didn't get to take very good notes. Some topics were
Since I kind of had to attend this one, missed 3 interesting talks. Tarik Keroum, from ObjectConnect (Smalltalk MT) on building COM/MTS systems in Smalltalk, Laurent Gauthier on XML and Smalltalk, and an experience report from Bell Atlantic, a big VisualWorks shop.
"I've been working on the VisualWorks 5I release, with a talented and wonderfully supportive team of enterprising human beings, which is a whole lot better than working with enterprise java beans" - Eliot Miranda's bio
Today's distraction: St. Patrick's day in Times Square.
Object People extravaganza morning. Anthony and Chulaka presented the work they did at Revenue Canada on thin-clients with transactions using Gemstone. Very cool stuff, but I figured I could get it first hand. Joel and Bob Huard presented their Bank of Canada experience report, also some very interesting stuff. Those slides should be available on our web site (http://www.objectpeople.com) soon.
One talk I really would have liked to see is the Hewitt Associates experience report keynote. They are "the largest multi-service outsourcer in the U.S." and mostly deal with administering benefits, human resources and compensation systems. Their system is implemented in VA Smalltalk, running on both client machines and IBM mainframes, and has been extremely successful, currently managing 50 big clients and 14 million participants. They are also featured in a full-page ad IBM had in JOOP to coincide with the conference. Slides of this keynote are available at http://www.software.ibm.com/ad/smalltalk/downloads/stsols99keynote.ppt. Quite interesting in terms of lessons learned, how mainframes work, and the VA Smalltalk capabilities there.
Instead of going to either of these, I spent the morning working on the ENVY book with Joseph and Jan. We had breakfast in the restaurant, then moved to the lounge until they kicked us out, so we went back to the restaurant and had lunch. Got some good work done, and was ready for the afternoon marathon.
Wednesday Afternoon's schedule was very intense, and no matter what, I was missing something I really wanted to go to. For example,
4:15-5:45
"A global mark-compact garbage collector is very fast when your heap is around 8K" - Andrew Brault
Pocket Smalltalk is an open source, cross-development environment with a 23K virtual machine and an 18K base class library. A 'hello world' app runs in 40K "ROM" 2K RAM. I wrote a simple demo app with a couple of widgets that takes 55K, but that's including the debugger. (Yes, I bought a Palm Pilot expressly for the purpose of writing this stuff).
The Pilot (sorry, Palm Computing Platform for the Enterprise) is a very odd environment. It has between 1-4 MB of RAM. Apps are typically very small. If you take more than 100K you'd better have a very good reason. There is no multi-threading. If the user switches applications then the current application has to shut down and save all its current state in non-program memory, and it has to do it fast.
Pocket Smalltalk lets you write pretty normal Smalltalk programs. They do impose some runtime restrictions, but remarkably few
They do go through a packaging process. It seems to take about half a second on my Toshiba 8000. The optimizing version seems to take as much as 2 seconds. Still a lot of rough edges on the environment, but it's open source. If you don't like it, fix it.
The other cheap Smalltalk that runs on small devices is Squeak. That doesn't have the same separation of development and delivery, so Squeak ports to handhelds are normally full-blown Squeak. That's cooler in some ways, but means their minimum memory footprint is on the order of 800K-1MB. By comparison, on Windows, an absolute minimal VisualAge Smalltalk app is about 2.1MB. An absolute minimal Java app is about 2.5 MB.
This thing is quite incredible. It was written by one guy, Andrew Brault, who just turned 21.
For more information, see
"I'm not allowed to use the name Java because I'm not a Java licensee, but Sun can sue me all they want, because I have no money" - Eric Arseneau
Eric Arseneau has been writing Pocket Java. This is a Java implementation that runs on top of the Pocket Smalltalk VM. It runs pretty directly, as he only had to add 2 special bytecodes to make it work. His implementation reads in .class files and spits out Smalltalk bytecodes for them. All he really does is run as a packager. It eliminates duplicated constants in .class files (which there are a lot of, he compressed 75,000 literals in the base libraries into 10,000 just by removing duplicates) removes the class file header (64 bytes per class). Another optimization knocked out 80K by changing the initialization code for Character. Java has no literal arrays, so they have a method explicitly initializes 4 arrays of 8,000 elements each. He deals with namespaces by replacing all occurrences of "/" or "." with "_". He ignores some Java semantics, for example, int overflow. This required him to change hashtable code, since Java relies on overflow to make them work. As you can see, he's not exactly honouring standard Java semantics. On the other hand, he fits hello world into 54K memory, including VM.
One interesting aspect was that he didn't have to worry about a lot of the tricky Java semantics, because he's a runtime, not a compiler. For example, static methods are represented as class methods. How does it deal with the fact that these can't be inherited. Simple, the compiler won't have let you write code that calls them on a subclass. The runtime doesn't care.
http://www.pocketjava.com is not active yet, but watch the space.
"We have not lost, and anyone who comes to me with that attitude can kiss my ever-lovin' Smalltalk ass" -Eric Clayberg
This was a panel session, moderated by Eric Clayberg, and featuring
I'll give an overview of the highlights.
"Out your employer!" - Eric Clayberg
Eric was the moderator, so he didn't say a whole lot. Some good points.
We are continuing to do well. (aside: Talking in private to people it's clear there is still an enormous amount of work going on, and although things are not at the insane pitch of a couple of years back "I used to quit one job and have another one the next day" fears of imminent unemployment are greatly exaggerated. Smalltalk Systems, for example, is selling 4 times the volume that they were last year, and Silvermark has done a VisualWorks port of their "Test Mentor" which is making them a lot of money (the great thing about a testing tool is that it can be brought in very late in a project's life cycle much more easily than something like TOPLink))
There are now 9, count-em, 9 different commercial Smalltalk vendors.
not even to mention the non-commercials, including
Many of those don't show up on the radar of high-end training and consulting companies, but they certainly exist and are out there in use by real people.
What can we as a community do to help support Smalltalk?
Eric's slides are available at http://www.smalltalksystems.com
Ed gave a measured, very technical presentation comparing Smalltalk and Java productivity and costs. A lot of it is based on numbers from Software Productivity Research, an organization founded by Capers Jones, which does a lot of software metrics work. One of the things they rate, using real, large projects as examples, is productivity in terms of lines of code per function point. They rate a whole lot of languages this way, but three of the more interesting numbers are (smaller is better)
This was quite surprising to me, as I would have expected a significant productivity gain from Java. Maybe I'm biased, but I would have thought that the garbage collector alone would be a factor of 10,000. Part of the reason is given as the immaturity of the Java tools (I don't know why that affects a LOC measurement) and the relative lack of big projects to measure.
Another interesting number was errors per function point.
These numbers are also consistent with something Mike Norman was telling me, about a company that did a (relatively small) project with two separate teams in both C++ and Java. The Java team took longer, mostly attributed to unfamiliarity with the language, but had lower defect rates at the end.
Ed had also done comparisons against projects of his own, and against the tables in the Microsoft Press Book "Rapid Development" (which is a very good book) which agreed with these numbers. He then went on to use translations from these numbers into time and cost values that made a rather good case for why you might want to go with Smalltalk.
Slides from his talk will be available soon at http://www.lineaengineering.com. Right now there's a very condensed version that appeared as a VisualAge magazine editorial.
"People keep asking me, what happens if IBM drops Smalltalk? IBM is not dropping Smalltalk, OK. Get over it" - Peter Spung
Here's Peter's write-up from the IBM web site.
I said it was easy to be a Smalltalk advocate because of the importance of it to IBM's customers, business partners, and IBM itself. I described some of the eye-popping systems that are running entire businesses for companies large and small: a stock transfer system that is in production and scaling up to handle 50% of the trades and 250,000 end users in North America by the end of this year; several core call centers for order-taking and customer support that support 7,000 to 10,000 users; server-based systems that handle two million financial transactions per day and require 24 x 7 x 365 operation. Nearly 400 add-on products are available for VisualAge Smalltalk from 325 business partners, and another 800+ upgrades and new products are planned. All totaled, I estimated that the annual industry around VisualAge Smalltalk is $50 million dollars and that around Smalltalk in general 2.5 times that, or $125 million, making it a viable and important marketplace. I stressed -- and will continue to stress -- the point that IBM is in Smalltalk for the long haul, as evidenced by its use within the VisualAge product line itself, the core frameworks and industry solutions in manufacturing, insurance, banking and finance delivered by IBM using Smalltalk, and the reliability and support expectations of products like VisualAge Smalltalk Server for the AS/400 and OS/390.
That's 50% of all stock trades in North America, at DST Systems, with 250,000 users. He also stressed that IBM does not port products to their mainframe systems lightly. That represents a major investment, which in IBM terms means hundreds of millions of dollars for them and for their customers. IBM has also recently rekindled their university program for Smalltalk.
A legitimate question is, if IBM is so committed to Smalltalk, why is there so little marketing in comparison to Java. Asking that question to various people, I learned a lot, most of which is not repeatable, about how large companies work. Within IBM, the marketing reporting structure is completely different from the development reporting structure. Commitments in terms of development don't seem to be strongly correlated to commitments in terms of marketing, and of course sales is its own thing altogether. What seems to be happening now is that WebSphere is the current big marketing push for IBM, and both VA Java and VA Smalltalk are trying to coordinate their releases and integrate with WebSphere in order to take advantage of that push.
One point Peter made was that in advocacy it's not sufficient to just bash Java. "Let's have first class interoperability with Java and hook to all the plumbing they're doing."
For more on Peter's impression of the conference see http://www.software.ibm.com/ad/smalltalk/events/stsols99rep1.html. It also has some interesting information on the Ultra-Thin Client and mentions other forthcoming IBM features, like Smalltalk stored procedures for DB2.
"Mean Time Between Failures of hours is simply not acceptable" - Eliot, quoting from a review of Java 'Enterprise' systems
Eliot Miranda, unfortunately lost his slides and had to fall back on generic "Why Smalltalk is good". In essence that it is a programming language focused on human beings rather than on the computer. Features like blocks, keyword arguments, and readable syntax are very significant. It's also a technology that's mature, with ideas that have evolved over 30 years and stable VMs and libraries that have been in use in mission-critical applications for more than 15 years. Reflection is extremely important, and in the business environment we are moving to, where most applications are 24x7, static typing simply doesn't scale.
"Eventually, in the business world, those that ship win" - Randy Best
Randy Best gave an overview of STIC's role, and some of the activities they're undertaking. If you haven't visited the STIC web site you absolutely should. He also outlined some of the steps the industry needs to take. He says that IBM should declare Smalltalk a mature, supported technology, that the ANSI committee needs to reconvene to consider some additional issues, particularly namespaces. He also urged people to refrain from just bashing Java and other non-Smalltalk languages, preferring to focus on the positive. He plans to provide computing editors with large amounts of information, and encourages people to help out. The STIC report on the conference is available at http://www.stic.org/Adventure/B020.htm. In his words
Smalltalk lives on and continues to do the hard stuff in application crucibles around the world. Remember when pundits told us COBOL was dead? How about the dire predictions in 1992 that mainframe computers were dead? The truth is out there, and 1999 will reveal this truth! Professional software developers use the tools that work best for the tasks at hand. Corollary: that which works, works. Support your Smalltalk vendors and continue delivering high-quality Smalltalk applications. In the end, those who "ship" win!
This year, Smalltalk Solutions featured a couple of evening "dinner" tutorials running from 6:00pm to 8:00pm. Given that it's been a long day already, I'm on expenses in New York City, and the dinner they're offering is cold and boxed, I'd have been very tempted to skip this session. Unfortunately, I'm presenting it, along with Joseph Pelrine, so it probably would have been a bad idea.
This tutorial consisted of a lot of preview material from the book. Slides are available on the web from http://www.daedalos.com/~j_pelrine (the web page is an experience in itself) and, coming soon, from http://www.objectpeople.com. Code, including a usable integration of the Refactoring Browser into both VisualAge and VisualWorks, and a three-way differences browser for ENVY, also available now at http://www.daedalos.com/~j_pelrine.
"Maybe objects are a bad idea, and building new abstractions does not have business value. It's possible" - Kent Beck
It's not over yet. Today has full-day tutorials scheduled, including John Lord on web development with VisualAge and Joe Winchester on the inner secrets of Visual Programming. Both are quite interesting, and I hop back and forth between them for the morning, but my brain is full, and I skip out for the afternoon and wander around New York. The technical content has been outstanding, overall, and I need to collect slides for some of the things I've missed.
The conference has inspired me to get out of doing any more Java, so of course I get to spend half an hour in an entertaining conference call regarding TOPLink/Java EJB support. I get to fly to San Francisco with Dan MacKinnon for the entire following week to deal with this stuff. Dan's a nice guy, but doesn't really compare to being home (and I'm sure he feels the same about me). The whole thing is kind of like selling your soul to the devil and then finding out he's holding up the paperwork. Our hotel in San Francisco will host a convention of pathologists.
"I know I can't be the only... whatever I am in the room" - Ani di Franco
The number of paid attendees this year was definitely below what SIGS needs to make a profit, so it would obviously be unwise for them to do the same thing next year. On the other hand, in some ways this is a rather anachronistic conference. Sure, it's nice to have our own exclusive conference and hold it in the most expensive place we can find. On the downside, that's kind of a holdover of the days when NY investment banks were the big Smalltalk users, money was no objects, and midtown Manhattan was a convenient location. It also makes it very hard to attract users. If someone's interested in finding out about Smalltalk, step one is to shell out US$5K to attend a conference, rather than just walking into a session. ST Solutions also competes with a lot of other conferences, including SIGS Object Expo (or whatever it's called now) held in Manhattan, two weeks later.
So, what's happening next year (and I'm not just making this up, this is what we were told by SIGS) is that Smalltalk Solutions will be folded into Object Expo NY as one or two tracks of Smalltalk material. This something SIGS does a lot, changing titles and re-arranging, so next year will probably be something like Application Development 2000 New York, incorporating Object Expo / SIGS Expo for Java / Smalltalk Solutions. For a conference that started out with the assumption that we'd have to find another venue next year this one turned out pretty damn upbeat.
See you next year.