Smalltalk Tidbits, Industry Rants

Patrick Logan points to Ted Leung talking about generics in Java:

As I listened to the generics discussion, my mind was thinking something like this:

• Generics extends the amount of static type checking that you can do.
• Static typing checking is good because you make fewer mistakes.
• But wait, now you have to worry about wildcards and type bounds.

It seems that for average programmers this could be more work than writing the code itself. So now you can make a mistake in writing the specification (type) that is supposed to keep you from making a mistake. Doesn't sound like a bargain to me.

All that just to send a message you know is implemented in the set of possible recipients. In Smalltalk, we don't even need specific generic support - because we do that sort of thing naturally. The stiffly typed crowd tie their knots tighter and tighter, don't they.

Ron Jeffries implements bowling in Smalltalk, demonstrating TDD along the way. There's related commentary here

If you use Smalltalk, you just put a halt right there in the SSP page - into the normal Smalltalk debugger you go. Over in ASP Land, it's a trifle more complicated than that. But hey, rebooting every few minutes probably aids your productivity, right?

Smalltalk, on the page: <% self halt. %>

or, over in ASP-Land:

Here's the drill. What I discovered yesterday while frustrated enough to physically throw my laptop out the nearest window.

1. Start debugger.
2. Wait until the process load starts, then HANGS.
3. Open Task Manager.
4. KILL aspnet_wp.exe.
5. New aspnet_wp.exe starts.
6. Process continues to load in debugger.

ICK! I had this problem while working at Paramount. I never did figure out the real fix for it. This will work for a couple of debugging sessions, then you get to reboot the machine.

And they call this the future... You could also ponder Seaside...

An interesting post appeared on the VSE (Visual Smalltalk Enterprise) mailing list recently; here's the description:

The original Visual Smalltalk dialect by Digitalk has survived ten years of proclaimed discontinuity. Paradoxically, the languishing commercial interest in this legitimate successor of the Smalltalk lore had a positive consequence: it protected the product from sophistication. After stripping out the over dimensioned Envy-like Enterprise development environment, and forgetting the PARTS that nobody seemed to use in real projects, we are left with a simple, powerful, compact and fast Smalltalk implementation. Even though Visual Smalltalk does certainly have a prominent history, it is neither perfect nor finished. As it is usually the case with every corpus of fruitful intellectual work, there are still many interesting details waiting to be completed. To begin with, there is one crucial problem in Visual Smalltalk: it has a bunch of hidden classes and methods that nobody has disclosed so far. This document is an attempt to repair that historical mistake. It explains how to recuperate all hidden classes and how to retrieve all missing Smalltalk source code.

The mail came from Leandro Caniglia; here's a link to the document he posted. It's a Word document, about 86k.

Chris Petrilli explains why the static typing thoughts that some in the Python community are having are a very bad idea.

The ArcterJournal lets slip a great rant on the current state of IE, and where that leaves developers:

Lets be honest here. There won't be any new updates to IE for a very long time will there? It's been said so many times, the IE renderer supports a billion languages, versions, and testing it in the wiz-bang microsoft test facility takes years for a single change. Are the changes required to give proper standards support coming anytime soon? I really doubt it, and that makes me sad, because it causes more work for me, the people I work with, and makes our products not nearly as good as they could be. Hell, even saying "support only IE" won't work because writing code for IE 6 (windows) still varies greatly compared to IE 5.X (latest on OS/9 and OS/X). You guys almost did us a favor by not supporting IE on the Mac anymore, which would have meant we could write for IE6/Mozilla, which would have been a bit easier, but later reversed the decision and re-supported IE 5 on the Mac.

So I ask again, how is the IE team working to make sure their browser is the best?

There's also that helpful way they decided to stop supporting the old Netscape plugin API, and instead decided to invent a new Active/X based scheme - breaking a bunch of code and lowering security in one fell swoop. Sigh...

In a post on the value of TDD, Sean McGrath says this about dynamic languages on the stiff VMs:

Jython : Anybody coding in a statically compiled programming language that is not using a dynamically typed programming language for the tests, has missed a golden opportunity for a productivity leap. In the C world, CPython. In the .NET world, IronPython. In the Java world, Jython.

If your productivity goes up that much, why limit those languages to tests? Why not just ditch the stiff language altogether and get a leg up on your competition?

Apparently, you can only give your customers (and prospects) the middle finger for so long - even the really dim management above SCO seems to have started to learn something:

Shares in Utah's SCO Group went into a tailspin late Tuesday as news spread of both deepening losses and an apparent coup at the software company's corporate parent, the Canopy Group.

SCO shares closed at $4.51 in regular trading on the Nasdaq Stock Market, down 33 cents, or 7 percent. Then came SCO's dismal earnings reports for the fourth quarter and fiscal 2004; within minutes shares plunged another 46 cents, or 10 percent, in after-hours trading, to $4.05.

SCO, embroiled in multibillion-dollar federal litigation against IBM and others over its purported rights to the Unix and Linux operating systems, more than quadrupled its fourth-quarter losses. For the quarter ending Oct. 31, SCO's loss sank to $6.5 million, or 37 cents a share; the company had lost $1.6 million, or 12 cents, in the same period last year.

Investors already were absorbing news, leaked out in bits and pieces earlier Tuesday, about an apparent weekend coup that ousted Ralph Yarro, Canopy's longtime president, chairman and chief executive, along with Chief Financial Officer Darcy Mott.

Secretaries at Canopy's Lindon headquarters confirmed that Yarro and Mott were "no longer with the company."

I guess even the really slow learners over there have finally started to recognize where the lawsuit madness has been leading...

Our lead VM Engineer, Eliot Miranda, has pushed out a few details on the 64 bit work the team is doing:

The 64-bit System: Overview

---

The 64-bit implementation uses full 64-bit addresses for objects, providing the ability to fill the entire available address space with objects. For an idea of the effective limits today consider that current AMD x86-64 chipsets support a 48-bit virtual address and a 40-bit physical address, while average 64-bit object size is around 64 bytes. So the maximum number of objects here-on is theoretically (2 raisedTo: 40) / 64.0, or around 16 Giga objects in a system with a terabyte of memory.

The 64-bit System: Implementation

---

There are only three tagged types, 61-bit 2's complement SmallIntegers, 61-bit unsigned Characters and a 61-bit SmallDouble, subset of the 64-bit IEEE double-precision format that provides the central 1/8th of the IEEE range at full precision. The immediate floating-point format provides a very usable range (approximately -1.0d77 to 1.0d77) which overflows to full 64-bit boxed Doubles when results don't fit. It provides a faster and much more space-affordable floating-point, being about 2.5 times faster (about 2.5 times slower than SmallInteger arithmetic) and having no space overhead.

The rationale behind using only three of the possible seven immediate tag patterns is two-fold. First, measurements show that not much space is saved by doing things like packing seven byte symbols into immediates because intrinsically these short symbols don't take up much space anyway, and providing access to such packed immediate types slows down the path for normal object access in things like the #at: primitives. Second, by using three tag types we can make the #isImmediate, #isSmallInteger and #isSmallDouble tests faster, since they need test only a single bit, these tests being performance-critical to inlined arithmetic and object access in the various #at: and #at:put: primitives.

The 64-bit object representation is relatively more compact than the 32-bit one, resulting in only a 33% growth in object header size from 12 bytes to 16. In particular object headers no longer reference their class object directly but instead include a 20-bit "class index" that is used in all in-line cache tests and in object instantiation. Classes are held in a sparse table and accessed by dereferencing the 20-bit index. This saves 44 bits while imposing a restriction on total number of classes unlikely to be a problem to contemporary applications. Class objects are accessed quite rarely, for example when a message send fails to find a lookup in the method caches and has to do a full class hierarchy lookup, or when the programmer explicitly accesses the class via the Object>>#class primitive. 64-bit objects have a 20 bit identity hash field (and in fact the class index is the class's identity hash) giving 1 Meg hash values (up from 16383 in the 32-bit system) and a maximum of 1 Meg classes. The header of 64-bit pointer objects also includes the number of fixed fields (number of named instance variables) so that the accessing primitives #at: and #at:put: no longer have to indirect through the class to find how many named instance variables to skip over. Consequently array access performance is much improved.

We have gratefully adopted an idea by Mark Van Gulik to do with tagging objects. Because object headers comprise two 64-bit words they can be placed at an even or an odd modulo 128-bit boundary. In the 64-bit system PermSpace objects are on an odd boundary and OldSpace objects on the even one. This means that the store check is slightly faster, but much more importantly means that PermSpace can be placed anywhere in the address space. All that needs to happen is for a PermSpace segment to have its object table aligned on an odd modulo 128-bit boundary. This means we can implement shared PermSpace easily, allowing the operating system to dictate where to memory map a PermSpace segment. Thus bit 4 is a tag that distinguishes between PermSpace and OldSpace objects, and hence we call it "tagged perm".

In the 32-bit system we require all of PermSpace to be above all of OldSpace or vice verca. When shared perm was first implemented OldSpace growth was not implemented. It was therefore easy to implement shared PermSpace with the OS mapping it above all of OldSpace. But once we provided OldSpace growth and shrinkage by memory-mapping OldSpace segments it became much more difficult to guarantee that PermSpace is mapped above all other heap segments since the upper portion of the address space is typically where shared libraries, the stack segment and memory-mapping all collide in a manner best determined by the OS and not amenable to precise control from an application program. Mapping memory at low addresses is typically difficult because in the lower portion of the address space the C heap and the application's code and data itself collide. Hence we have yet to reimplement shared perm in the 32-bit system.

By using the "tagged perm" scheme we are able to decouple PermSpace from location and can hence easily share PermSpace, and allow PermSpace to grow. This first release preview release does not support shared PermSpace

As implied above, a preview (early beta) of the 64 bit VM is on the latest release CD.

Ambrai Smalltalk has reached another beta milestone - see the Ambrai website for details. Ambrai is a Mac specific Smalltalk implementation.

Thanks to Terry Raymond, there's an implementation of traits for VW now. From the page:

VW Traits is an implementation of Traits for VisualWorks. However, it is not a strict Traits implementation. Traits programming is a techinque of code composition and code reuse. A Trait is a partial implementation of a class, multiple traits are combined together, along with glue code, to comprise a class (client). You can view the original Trait papers here.

See the Wiki page for more details, or dive right in by loading up VWTraits and VWTraits Base VW Overrides from the public store

Sci Fi Wire reports that the sixth Harry Potter book (not outuntil July 16th) is already a best seller.

Imagine a street brawl involving a bunch of Santas - (via Bob Congdon)

Festive cheer and goodwill was in short supply in Newtown when people dressed as Santa were involved in a mass street brawl, say police.

Officers used CS spray and batons to break up trouble amongst up to 30 people, following Newtown's annual charity Santa run.

Can't say that I've ever seen a report that referred to Santa Claus and CS gas simultaneously...

Scott Knowles points to a Wired story that explains how the lack of scarcity really affects online marketplaces:

To get a sense of our true taste, unfiltered by the economics of scarcity, look at Rhapsody, a subscription-based streaming music service (owned by RealNetworks) that currently offers more than 735,000 tracks.

Chart Rhapsody's monthly statistics and you get a "power law" demand curve that looks much like any record store's, with huge appeal for the top tracks, tailing off quickly for less popular ones. But a really interesting thing happens once you dig below the top 40,000 tracks, which is about the amount of the fluid inventory (the albums carried that will eventually be sold) of the average real-world record store. Here, the Wal-Marts of the world go to zero - either they don't carry any more CDs, or the few potential local takers for such fringy fare never find it or never even enter the store.

The Rhapsody demand, however, keeps going. Not only is every one of Rhapsody's top 100,000 tracks streamed at least once each month, the same is true for its top 200,000, top 300,000, and top 400,000. As fast as Rhapsody adds tracks to its library, those songs find an audience, even if it's just a few people a month, somewhere in the country.

This is the Long Tail.

This is something that mass retailers are going to have to face. There's still a lot of pleasure to be gained in just browsing at a store - and you'll often find things that you wouldn't have noticed ootherwise that way (I spotted a bunch of history books at Borders yesterday that I'll want to get, for instance). On the other hand, search is pretty darn good at Amazon - and I've found things of interest that way as well - things that I would be unlikely to ever notice at a real store.

There's plenty of room for the big retailers in the mass market end of the business. What's opening up are the niche spaces that hardly anyone bothered servicing before. Your particular taste in music (or books, etc) might be rare enough that Borders would never bother to stock it - but the online retailer who has no actual inventory (beyond bits) can easily do so. As book publishing moves towards on demand binding, you'll see the same kind of disintermediation hit book sales that's been slamming music for the last few years.

Seems that it's not just me - The .NET Guy isn't pleased either:

I read posts like this one detailing the new generics system in .NET] and it makes the baby Jesus want to cry. Where is the latent typing that made generic programming so powerful? The monstrosity they've cooked up is basically useful for any type of generic programming where you don't actually USE the thing... meaning, containers. *sigh* I think it's time to start coding in IronPython.

Ted Neward hopes that XML will continue to get hammered as the uber answer:

XML will start to lose its luster. People are coming around to hate XSD Schema. Critics are popping up everywhere. Dave Megginson has a new book out, in fact, that basically takes everybody in the WS-* stack to task over creating specs before seeing if they'll actually work. In fact, the only successful application of XML thus far that anybody but a developer feels is weblogs and RSS, which obeys none of the classic rules of an XML spec. (Read his book, by the way--it's an eye-opener, particularly if you're one of the XML faithful, or if you've been thinking that XML will somehow make the computer world a safer, saner place. I'll put the Amazon link in here when I get a chance, but it's his latest from Addison-Wesley; shouldn't be too hard to find.) XML is useful, but we're starting to see the warts, and behave accordingly.

Hmm - where have I seen the spec a week without testing theory before? Yes, with the OMG, back in the 90's, when CORBA was all the rage. The WS* specs are a complete rehash of that entire silly process...

Michael Gartenberg seems to think big things are coming down the pike:

Just wrapped up another fascinating call. I'm still in thinking mode about the implications of what was discussed. The first few weeks in January are going to be very interesting. 2005 is shaping up to be a year of interesting products, relationships and technologies. Some of this stuff (if executed properly from a marketing and messaging perspective) is going to be amazing. You heard it here first :)

So is this another "big" announcement from MS (like Spaces), or something else?

No one is quite sure how this happening, but one of the Mars Rovers (Opportunity) is getting its solar panels cleaned off at night. Did Guiliani send all the squeegee men that far away? :)

An unexplained phenomenon akin to a space-borne car wash has boosted the performance of one of the two U.S. rovers probing the surface of Mars, New Scientist magazine said on Tuesday.

It said something -- or someone -- had regularly cleaned layers of dust from the solar panels of the Mars Opportunity vehicle while it was closed down during the Martian night.

The cleaning had boosted the panels' power output close to their maximum 900 watt-hours per day after at one stage dropping to 500 watt-hours because of the heavy Martian dirt.

"These exciting and unexplained cleaning events have kept Opportunity in really great shape," the magazine quoted NASA rover team leader Jim Erickson as saying.

Sounds like Enterprise is heading back to bogosity to me:

In the episodes, the Enterprise heads back to Earth for the official launch of the Columbia NX-02, Starfleet's second warp ship, commanded by Erika Hernandez. Phlox is abducted by aliens and finds himself in the presence of Klingons who tell him the Empire is facing its gravest threat in centuries. Along the way, as Archer and company investigate and pursue, it's revealed that one of our main characters has a secret past, which comes into play, the site reported.

The addition of Browder should work, so long as they don't let him write. No animated episodes!

I got a chuckle out of this touting of generics support on the CLR - scroll a bit more than halfway down and look at all the rules. Now, contrast that with what it takes to do the same thing in Smalltalk - just implement the darn method in the class you want it in. No fuss, no muss, no nine million rules to follow. I'm sure that the MS team thinks that those rules make things cosmically safer somehow; all they do is add bricks to the mental backload you have to cart around while writing software. Via Julia Lerman.

I can almost see the lines now :) Reuters reports that "Harry Potter and the Half-Blood Prince" will be out July 16. Can my daughter wait that long? Can I?

One of the persistent problems for newbies taking a look at Smalltalk has been the "where's my application?" problem. We haven't got a runtime system defined yet, but there's been progress in the 7.3 release of VisualWorks. New in this release is Subsystems work done by Alan. Here's an excerpt from the documentation:

It is frequently necessary to take special actions when certain system events occur, notably when the system starts up, shuts down, and immediately before and after an image save. The order in which such actions occur, relative to other parts of the system, can be critical. For example, a GUI application probably needs to perform and window startup routines only after the windowing system itself has been initialized.

Traditionally, startup events have been handled by registering dependencies on ObjectMemory. More recently, SystemEventInterest instances have been supported by the system. Both of these mechanisms left it difficult to manage the order in which actions were taken.

Class Subsystem provides VisualWorks a simple way to specify dependencies on system events as well as a modular approach to controling their order of execution. Several subsystems are defined for handling VisualWorks startup procedures.

Two subclasses in particular are of interest to the application developer:

• UserApplication
• ImageConfigurationSystem.

If an application has actions to perform upon one of the four system events, a subclass of UserApplication is a convenient place to specify those actions. ImageConfigurationSystem is useful for applications that process command line options.

Defining System Event Actions
Subsystem defines four system event messages to which subsystems can respond: activate, deactivate, pause, and resume. By default, these general events are invoked as follows:

• activate is invoked by #returnFromSnapshot, which occurs when an image is launched.
• deactive is invoked by #aboutToQuit, which occurs just before the image exits
• pause is invoked by #aboutToSnapshot, which occurs just prior to writing an image file
• resume is invoked by #finishedSnapshot, which occurs just after the image file has been written

Responding to System Events
Some subsystems invoke activate upon #earlySystemInstallation, but these are usually system level subsystems. For applications, #returnFromSnapshot is the appropriate system event. A subsystem does not respond to these system event messages directly. Instead, these messages invoke further messages in which a subsystem configures its response to the system events. The corresponding messages that a subsystem will implement as needed are:

• setUp
• Defines actions to perform upon the activate event message, and activates the subsystem.
• tearDown
• Defines actions to perform upon the deactivate event message, and deactivates the subsystem.
• pauseAction
• Defines actions to perform upon the pause event message.
• resumeAction
• Defines actions to perform upon the resume event message.

An application seldom needs to perform actions before or after a snapshot, which is generally a development time activity, so do not generally have to provide implementations for pauseAction or resumeAction. An action does, however, frequently have actions to perform upon launching the image, such as setting up its runtime environment, and these are specified by an implementation of setUp. Less frequently, but not uncommonly, an application will also need to perform actions prior to shutdown, which can be implemented in the tearDown method. The UserApplication subsystem, which is intended to be the superclass for application subsystems, implements one additional stub method:

main
This method can be implemented by a subsystem to launch the application, as well as to perform other application set up tasks. This method simplifies starting an application upon image launch, eliminating the need to either save the image with the application open, or of using Runtime Packager to specify the application to run, or any of the other methods that have been used. As an example of using setUp and tearDown methods, consider the task of saving a random number seed upon shutdown and then reading that seed to restart a random number generator upon startup.

This is something I'll be migrating BottomFeeder to after the 3.8 release - which means that post 3.8, we'll be moving to a VW 7.3 runtime. I'll be posting comments on how I'm migrating to the new code - it should be useful information in general.

To see the full documentation on this stuff, open the Application Developer's Guide, and look at the Application Frameworks chapter.

The release of CST Fall '04 has hit the press wires:

• Yahoo UK
• Market Wire
• Yahoo Canada
• Yahoo Biz
• Lycos Finance
• Excite Money
• CNN Money
• MSNBC
• My Way Finance
• Hoovers
• Newstream
• Optionetics
• IWON Money
• Vantage Link
• American Chemical Society

Ok, this is interesting. Tim Bray states that they are having some "dynamic infrastructure" built into NetBeans:

So, here's what we're doing: The NetBeans group and the Software CTO Office (where I work) have pulled together a project to fix the problem. We're going to be paying David Strupl, a contractor who really knows NetBeans, to lead a java.net project to build dynamic-language infrastructure into NetBeans. Obviously it'll be Open-Source and everyone who wants to can play.

I have to admit, I have no idea what that means. The problem with dynamic languages on the JVM isn't really one that can be solved easily at the IDE level; it's more of an infrastructure issue down at the VM level. The flaws that I outlined in the CLR also exist in the JVM; it really, really wants a static language running on it. Sure, you can host a dynamic one there... so long as you are willing to accept a slow or compromised language. My question is - what the heck is "dynamic infrastructure" at the NetBeans level? What does that mean?

I think this counts a major oops:

A newly reported security problem in Microsoft's Internet Explorer Web browser allows attackers to create a fake Web site that looks exactly like a genuine site.

The vulnerability lets an attacker display any Web site while the address bar in IE will display a trusted Web address and even show the icon indicating SSL (Secure Socket Layer) security, security researchers warn.

Sheesh. How is an end user supposed to spot that? Time to load up Firefox...

There's a mini-tempest swirling over the use of the term beta as applied to software; Ryan Lowe tries to make sense of it in light of Mary Hodder's rant:

I look at this problem as a software engineer and the word "beta" seems as insignificant and arbitrary as version numbers. There's no universal quality measuring stick you must satisfy to get out of beta, it's just an arbitrary developer-defined state meaning "less than perfect". Well guess what? The final release will be less than perfect too. All beta means is that it's less than less than perfect.

That's the stumbling block that users need to get over when it comes to software. All software is imperfect and there's nothing software developers can do about it. It would be impractical to make software perfect because it would cost far too much and take too long to release. Just ask NASA how much work it takes to get defect-free software. Even after all of the checks and balances NASA has, they still have disasters caused by software. The solution for software companies is to release less than perfect software that does a pretty good job.

That's true enough - I've certainly let revs of BottomFeeder out with issues; half the reason I built the online updating feature was to allow for my own limitations :). Still, it's useful to look at what individual vendors mean by the term beta. In the case of Google, it seems to have no real meaning at all (at least, from the end user's perspective). In the case of NetNewsWire (the source of the problem that generated the original rant), take a look at what they say about their own beta software:

Warning!

Beta software has bugs! Nasty, vicious bugs with great big, sharp teeth!

Don't use beta software unless you're clear on what "beta" means and you're comfortable running beta software.

Seems to me that the guys responsible for NNW are pretty clear that you are taking risks running beta software. The term may have lost a lot of meaning due to rampant misuse, but not over there - I think they've been very clear. Ultimately, you need to back your data up. I don't do that enough, and I'm sure most people are like me in that regard...

I can't claim credit for the term; it was Alan's idea awhile back on the irc channel. However, one of the regulars (Rado) had a nice take on it:

The 3 laws of objections:

1. You must harm human beings, or allow through inaction a human being come to harm
2. You must reject all instructions without violating (1)
3. You must protect yourself at all costs

Heh

I added two new settings to BottomFeeder today:

• A setting that adds a variable delay between http updates during the update loop. It slows the update cycle down, but should reduce the cpu load during the update cycle
• A setting to manage the priority of the update process. You can set it to high (same as it has been up to now), medium, or low. Lowering the priority will improve UI reaction during the update cycle.

To get these, grab the latest dev update and restart.

I noticed something interesting about the wiki spamming (of the VW Wiki at uiuc and of the CST Wiki - it takes place Monday through Friday. During evening hours (to me, on the US east coast). This leads me to believe that the spamming of these Wikis isn't random - it's actually part of someone's job. Unbelievable.

I'm getting the distinct impression that the law arms of big firms have with the sales and marketing groups. I posted awhile back on this headache that Sun's lawyers were giving Ted Neward; now I read where MS' lawyers have been hassling MacSlash. In Ted's case, the marketing folks were properly horrified once they found out, and it looks like Scoble caught wind of MS' lawyer mess.

This points to a fundamental problem in the field though - legal departments are off chasing trademark, copyright (etc) problems down without regards to business issues. In the process, they are giving black eyes to their firms. Sounds to me like it's time for the marketing departments to start exercising oversight...

I've done some more fine tuning of the comment policy. If you look at the blogs here, you'll find that there are 2 policies at work for displaying posts:

• Display all posts from the last N days
• Display the last N posts

I use the first policy for blogs - like mine - that get posted to on a regular, daily basis. I use the second policy for blogs that don't see the same level of activity. As it happens, that policy affects comments. For my blog, a comment on a post from October is not really relevant - that post has long since scrolled off the page and into the archives, and I just have comments disabled for it. For the blogs with less frequent posts, it makes sense for comments to be enabled for any post that is on the main page - which is how it now works. In general, comments are disabled for posts that have walked off the current main page.

I've had BottomFeeder copying the current data files to a backup directory - so as to preserve a "known good" rev of the data before startup for awhile now. yesterday, I took a look at that code and realized that the code was making an invalid check - and thus could possibly overwrite a good backup with a blown save file. That's not good. So, I took a look at the startup sequence, had the typical who wrote this crap? Oh yeah, that would be me kind of reaction, and fixed it. The dev update has a much cleaner startup now, and it'll attempt to read the backup date if the normal save file is blown - and won't, in fact, overwrite the backup data in that circumstance. Should be cleaner and safer (although you should still backup the data yourself as well).

Mark Bernstein talks about the flap that came up recently over data loss - have a look here, but be prepared for a huge rant and set of counter-rants. The bottom line is that this person lost an impressive amount of data, and was upset about it. As Mark points out,

• All Software has bugs
• This was beta software with warnings

That's not the point that I really want to get to though. Take a look at Mark's last paragraph:

Software pricing is out of whack. In the instance, an influential professional was using free 30-day trials to select a professional tool she used extensively every day, a tool that was close to the center of her worklife. How much would the successful vendor receive for winning the competition? $24.95. This is absurd.

This is a general problem in the software field right now. People expect tools to be either free or cheap. Ponder for yourself what it takes to pay a few developers - plus the hardware and software (and office space) that they need to work. Now factor in the potential size of their market - and consider how many copies they would have to sell at $24.95 to even come close to the break even point. Now, if that software is mission critical....

In an interesting post on a customer visit, Jonathan Schwartz has this exchange on RedHat, Linux, and lock in:

And then I asked about IBM. And apparently they'd just been with OSDL, who'd evangelized that with open source, there was no lock in. When I pointed out that OSDL was led by a 17-year IBM veteran who should know better, the CIO started laughing as if I was joking. So I suggested they read the OSDL website , and revisit some software basics. IBM told him they couldn't get locked in with linux. And I said, "nice vision, but Red Hat has you locked already." The CIO shrugged, "nah, it's open source." My response, "Have you tried replacing what you're deploying?" He asked his lieutenant, who said "we can't get vendors to qualify to any distribution other than Red Hat. We don't have a choice. He's right." IBM, up to its old tricks again.

That's increasingly the case in the Java app server world (which is fascinating all by itself - isn't Java supposed to be x-platform?). Oddly, the biggest issues we have with Linux is on Red Hat itself. We've had library issues, and recently we've had problems reported on Fedora Core3. We've had nothing similar reported on other Linux distros. It would be interesting to know whether it's ineptness on their part, or part of a plan.

Doc Searls doesn't like some of the surreptitious marketing that's going on:

Several people have asked me what I think about BuzzAgent, so here's my joint answer to all of them: It sucks. Where Marqui is up front about what it's doing, and engaging bloggers in conversation (as well as promotion), BuzzAgent and its clients are being surreptitious and false, and spreading a virus of falsity through its agents. Mass market advertising has always been impersonal, and often (okay, almost always) fake. BuzzAgent's system allows advertisers to be no less fake, but in person, face to face. Even if the agents really do love the products they shill, their love is bought. Worse, it comes cheap.

To these kinds of marketers, "markets are conversations" means "delivering messages" through talk. What they miss is that the next stage beyond conversation is relationship. And that relationship isn't just with a "brand." To have real value, the relationsihp needs to be with the people behind that brand. And that relationship takes place in the public marketplace.

I fail to see the harm. People become unpaid advocates of products all the time; these people are simply moving one step up from that. Go see a movie; those products you see the characters using are all part of product placement - which is pretty much the same thing. Hang around a bunch of developers sometime and listen to the Java advocates arguing with the C+ developers, or with the Smalltalk developers. Are those people cheap shills too? Or are they somehow purer due to the fact that no one is rewarding them for their love of a particular product?

If this is true:

The Al Fresco campaign is over -- having notably boosted sales, by 100 percent in some stores -- but she is still spreading word of mouth about a variety of other products, and revealing her identity, she said, would undermine her effectiveness as an agent.

Then you better get used to it - because any campaign that works is one you'll see more of.

I think this qualifies as a major oops...

Microsoft this week quietly fixed a security weakness in the configuration of the built-in firewall component of Windows XP. The firewall - turned on by default by XP Service Pack 2 (SP2) - can leave files open across the whole net if users choose to enable file and printer sharing, it transpires. Such access should be restricted across a local network but Microsoft has implemented the feature in such a way that, for users of some dial-up ISPs, a local sub-net becomes the whole internet. Microsoft first informed users of this back in September but it has taken five months for it to release a fix.

This quote comes from an amusing source - it's a spam referral thing, but - oddly enough - it makes a point worth commenting on. I'm not going to link to it, since that will only encourage this kind of abuse, but - I've seen this argument before:

Some people argue that static methods are not object-oriented since they do have the semantics of a global function; with a static method you don't send a message to an object, since there's no this. This is probably a fair argument, and if you find yourself using a lot of static methods you should probably rethink your strategy. However, statics are pragmatic and there are times when you genuinely need them, so whether or not they are "proper OOP" should be left to the theoreticians. Indeed, even Smalltalk has the equivalent in its "class methods."

Nope. Smalltalk classes are actual objects, like everything else in the system. Class methods are actually instance methods - of the class. The class is simple the sole instance of its meta-class. As such, when I create Bar as a subclass of Foo, all the class methods are inherited. Want an example of this? Go find the inherited implementation of new - it's not, in fact, a class method in Object - rather, it's an instance method in Behavior...

In a comment on this post:

We all know that changing language is both risky and expensive. Imagine you are on your current, albeit non-optimum, path of progress -- up and to the right, but not as fast as you would like. Now look at switching to Smalltalk (or Lisp, or...). First you must stop your "up" progress while you learn. You are now behind where you would have been. You hope the Smalltalk (or...) vector is more "up" than your previous, but of course you do not know how much more -- just lots of opinions from various people you do not know (and therefore trust). At sometime in the future -- is this 3 months or is it 12 months? -- you finally catch up to where you would have been had you not stopped to learn. Only now do you begin to reap the benefits.

That's not at all the scenario I sketched, and not at all the place Yahoo was in. I would almost never advise a technology migration (there are exceptions for completely out of date/out of business solutions). In the case I brought up, Yahoo had a working system in Lisp, and migrated it to C++ because some thundering moron decided that it would be too hard to find/train Lisp developers. So instead, he spent huge sums of cash to build a less functional version of what he had to start with - most likely with a bigger team. You know what that's called? Stupidity, plain and simple.

Sure, if you have an existing system, it makes very little sense to migrate it. Which is why all the frenzied "rewrite it Java" activity of the last 7 years has been so utterly asinine...

There's a Smalltalk Meetup in January - January 20th to be specific. I'll be attending - and handoing out Cincom Smalltalk non-commercial CD's. Here are the details:

Starbuck's Coffee (Chinatown)
800 7th St., NW
Washington, DC 20001
(202) 289-1576
Map

See you there!

I accidentally turned comments off earlier today; it should be back to normal now. This only affected people who comment via the html page; if you use an aggregator (and thus, the CommentAPI), it was never an issue.

Update: Comments actually work again now. It's a bad thing to forget the false half of a true/false test :)

This post about a catastrophic data loss with NetNewsWire makes something very clear to me - backups are important. I know that people have lost data with BottomFeeder. I try to be careful about user data, but bad stuff does, unfortunately, happen. Here's what you need to backup from Bf - the contents of the btfSave directory. That's where all application data resides, so if you back that up, you should be able to get right back to the saved state easily.

We released Cincom Smalltalk Fall, 2004 a few days ago - and orders will start going out before the end of the year - December 20th is when the actual orders will go out here in the US (possibly earlier for our overseas offices). You should receive the new CD's before the New Year.

Ted Neward points out that security issues are not only a Microsoft problem:

Microsoft isn't as much the hotbed of vulnerabilities as you might think; yes, they have their fair share, but the key words there are "fair share", not "lion's share" or "biggest burden" or "crappy software". What makes their vulnerabilities so dangerous is the ubiquity of their software, not the quantity of the vulnerabilities themselves.

Of course, the number of security issues that affect MS software don't tell the whole story - as Ted points out above, and as I pointed out here, the ecosystem of the internet makes any MS vulnerability orders of magnitude worse than an equivalent problem for Apple, Sun (etc). In this sense, MS is a victim of their own success - and as a result, they have to just suck it up. Unlike Ted (read his whole post), I don't think SP2 really qualifies as sucking it up though...

Ian Bicking makes some points about why Lisp (and by extension, other niche languages) don't always make it in the market:

Lisp makes good programmers really productive, more than they could be in another language. Paul Graham talks about this in Beating the Averages. He made great software and sold it for a bundle to Yahoo. But now it's been reimplemented in C++_. Why, oh why?

It's easy to blame stupid people for this sort of thing, except that it keeps happening over and over. Metaprogramming is powerful, and was central to Viaweb (20-25% of the code, according to Graham). I think this is an example of Common Lisp's fatal flaw (and since Common Lisp is the standard bearer for all Lisps, it is Lisp's fatal flaw).

Some of what Ian points out is true - metaprogramming is hard, and making developers confront it is often times a problem. However, I really, really doubt that it had a lot to do with Yahoo re-implementing in C++. I'd warrant that 90% of that decision was based on fear. Fear that:

• He wouldn't be able to find Lisp developers
• The Lisp developers he found would be more expensive

There's also tons of weak thinking going on in that kind of management decision - the concept of training a developer never seems to cross management's mind in these situations. The theory seems to be "well, training our developers in blah would be expensive (in which case, it never made any sense at all to move to Java or C# - or even C back in the day).

Managers make development decisions in much the same way that people approach lines - they engage in herd behavior. Ever notice how people tend to cluster at tollbooth lines on the highway? It's the same thinking that drives the decision to use (insert fad language of the moment here). It's a relatively rare individual who can really make a break from herd thinking and take the road less traveled.

The way I like to describe this is as follows: If you decide to go with the same tools and technology as everyone else, you make sure that you won't fail any worse than they do. However, you also ensure that you won't succeed any better. Mitigating the risk of failure is enough for most people - risking something "unknown" for a chance at better success sounds too risky - which is likely why Yahoo reimplemented the Lisp system in C++ - if pushed the system into that manager's comfort zone.

Martin Fowler brings up the problem of using metaphors to describe software development - his point applies to any metaphor though - you can only bulld metaphorical bridges so far before they end up over open water:

As regular readers of my work may know, I'm very suspicious of using metaphors of other professions to reason about software development. In particular, I believe the engineering metaphor has done our profession damage - in that it has encouraged the notion of separating design from construction.

As I was hanging around our London office, this issue came up in the context of Lean Manufacturing, a metaphor that's used quite often in agile circles - particularly by the Poppendiecks. If I don't like metaphoric reasoning from civil engineering, do I like it more from lean manufacturing?

I think the same dangers apply, but it all comes down to how you use the metaphor. Comparing to another activity is useful if it helps you formulate questions, it's dangerous when you use it to justify answers.

So as an example - one of the principles of lean manufacturing is the elimination of inventory. This leads to the question of whether there is an analogous item to inventory in software development.

As soon as I read that, I realized that I had heard the term inventory applied to software development before - I finally realized that it was in Scott Ambler's talk at XP Brazil a couple of years back:

Metaphor from this morning - the data (dba) community is packing for Antarctica. Unfortunately, the rest of us are traveling in the desert or the jungle

Scott used that in describing what kinds of things you put in a backpack - depending on what kind of hike you are planning to take. So going back to Martin's question - is that a useful metaphor? Well, it does make you ask questions along the lines of will we need it? In that sense, I guess it's useful. The difficulty is, the answer to that question is going to differ based on who you ask. Like anything else, whether you need a given thing (up front documentation being Martin's example) is going to be a matter of opinion. Software development is still a field where you need to apply consensus rules a lot.

I noticed earlier today that the comment spammers seemed to be concentrating on older posts (the theory being that I wouldn't notice, I guess). The upshot is this - comments will be rejected for any post older than 4 days. There was someone trying to comment on this item that got slapped by that; it's another case where the a**holes have ruined it for the whole class :(

Runar Jordahl has some biting commentary on the MS purchase of an anti-spyware company

We released CST Fall 2004 edition earlier this week - now the non-commercial is ready to be downloaded. Point your browsers at our download page and grab it. If you've already registered, just login here. Enjoy!

Today is just a great day for a power outage - it's a nice balmy 20 something outside, and bam - off it went. At least my battery backup gave me time to do a proper shutdown of the Linux server. I was wondering what the funny clicking sound was - apparently, the power flow was getting all wonky just before it went down. It would really, really suck to have a cold house and lose all the stuff in the freezer... Only down for about 90 minutes, so I guess the freezer is ok. I suppose cleaning up my office has some value...

I had been completely bored by ipv6 until I read this InfoWorld piece:

Recently, cows in Gifu prefecture were tagged with tiny networked devices to wirelessly track their movements and body temperatures for health and breeding purposes. And in Nagoya City, taxis were fitted with Internet-enabled sensors on their windshield wipers, allowing dispatchers to continuously monitor rainfall via wiper speed and to dispatch more cabs to the wetter neighborhoods.

There's a whole class of new presence applications right there.

Just how bad is the spam problem for blogs getting? Have a look at this post from one of the MT folks. It seems that some MT sites are getting bogged due to attempted spamming - the cpu load of detecting spam is becoming a problem all by itself. It sounds like they have a handle on the problem, and have some fixes coming - but it's not their fault. This just gets uglier and uglier

The RSS perma-thread on resource consumption now has its very own blog.

Christopher Petrilli is not impressed with Solaris on x86:

Yup, one of the single most popular gigabit ethernet chips ever released is not supported. Sun supports, in total, 12 network adapters, and once you filter out the duplicate chipsets and such, it's more like 10. Ten. Ten. Last I checked, FreeBSD supported about 100+, and Linux a comparable number.

Do these people really think they'll displace anyone in the x86 world? Crapy install process, filled with confusing questions and useless prompts, only to get to a fully installed system to find out that in fact your HW isn't supported. That's a fine bit of code for you.

Maybe there's something amazing inside Solaris 10, but I'll never know since I won't fork over stupid amounts of money for a Sun box. And please don't point me at the Sun Blade 100, as I've had one, it's insulting slow. 386/33 running BSDI slow. For $1k, I expect at least passable.

So long, farewell, don't let the door hit your ass on the way out.

Kind of puts this in perspective, doesn't it?.

Update: Andrew Binstock of SD Times makes some good points about Solaris vis-a-vis Linux, but - and this is critical - the negatives that Christopher outlined above are of far greater importance. If you can't get the system installed, or once it's installed it can't see the network - it's useless