Apr 93 Editorial
|Column Tag:||The Editor's Page
MADA Conference Who, Where and What
By Neil Ticktin, Editor-in-Chief
For those of you who dont know, MADA is the MacApp Developers Association. Today, that is a bit of a misnomer - they are beginning to take on more than MacApp. For years, MADA has been touting (along with Apple) the benefits of object-oriented programming (OOP). If you havent at least heard of OOP by now, youve probably been hiding out in Northern Greenland waiting for the ice to melt.
For the past several years, MADA has held their annual MADACON conference. This year, the conference was held in San Diego, February 15-19 and had a few fireworks to go with it. Heres some of the happenings
There were a few very impressive things shown at MADACON. Kurt Schmucker from Apples Advanced Technology Group (ATG) gave the keynote on OODL Authoring Environments. The heart of his talk was about a technology nicknamed SK8.
First, I should make clear to you that they were not announcing product, just showing off a technology. For a while now, weve all been saying that ultimately computers need to be programmable by non-programmers. When Bill Atkinson first came out with HyperCard, his vision was to have a software erector set. HyperCard was a start, but serious developers never took it seriously.
It is clear that at this time, the software development market has a rich set of 70s and 80s tools, but a breakthrough is needed. ATGs task was to combine the ease of HyperCard with the power of MacApp. ATG had a few goals: leverage developers time, dramatically increase the developer base, and create the ability to develop for different platforms as well as different metaphors.
ATG soon realized that more dynamic languages (such as Dylan) were the better route. But, Dylan is not for everyone - they needed Object Oriented Development Languages (OODLs) for the masses. Their result: Object Oriented Dynamic Authoring Languages - OODALs for short. The key here is to empower subject matter experts to create content with less dependence on programmers.
The result of all this research is a powerful OODL Scripting Language with powerful graphics, direct manipulation user interface, and knowledge representation, inferencing, and constraint solvers. In other words SK8.
So what are the specifics? SK8 is built on top of Macintosh Common Lisp (which says a lot for MCL). A language called SK8Script is built on top of a group of systems - objects, events, 2D graphics, OO rules, and constraints. A set of core objects are then built in SK8Script. SK8Script and this set of core objects are what the author interacts with.
SK8Script has greater expressive power than HyperTalk, is simpler than HyperTalk, has all the power of MCL or Dylan, but is incredibly easy to read and write. There isnt room for a lot of examples here, but lets look at one. Lets say that you want to remove all the bold words in a document. To do that you would write:
Remove every word whose style is "bold" in the document named "foo"
Pretty simple stuff, but its powerful.
SK8 can be used in high-end multimedia authoring, vertical tools generation (i.e., online training systems and documentation systems). Today, SK8 is being used in conjunction with people at Boeing to develop a training system.
What does this mean to you? ATG has been teased for only working on technologies, but not actually shipping product. They are aiming to get some (or all) of the concepts worked into development platforms. Some people have termed SK8 as HyperCard on steroids. Hopefully, the HyperCard team (especially now that Apple has control of it again) will embrace some of SK8s concepts and integrate them into the next version of HyperCard.
Little by little, Symantec and Apple are letting out information about Bedrock. Some people think that the Bedrock team is holding back on information because it doesnt know a lot of answers yet. Even if this is the case, it is true that the Bedrock team (BT) is making progress.
At MADACON, the BT let us in on some specifics. Bedrock is an application framework, not just a GUI class library. It will be easier to use than the native API and is platform independent.
As of MADACON, Bedrock had almost 400 pre-tested classes and the quantity is continuing to grow. File handling will be platform independent and will take on the native look of the platform you are running on. Strings will be supported through ANSI string classes, validators (time, date, numerical), and formatters (for date, time, and currency).
The BT announced that they have licensed a subset of Rationals Booch components (maps, stacks, queues, etc ). This is great because Booch is a noted industry standard. Their tools are second generation and they therefore know a lot more than most on the market. As a result, the Booch components are flexible and complete.
The design philosophy behind the data structures is seemingly well thought out. They are template based with mixin inheritance, exceptions, public inheritance for interface, and private inheritance for implementation. Bedrock will provide tools for sequential and binary searches, QuickSorts, and storage allocators.
Jim Lloyd took a moment to discuss exception handling and why they took that path. Exceptions are a good error handling methodology, and furthermore, allow for the use of multiple independent libraries (e.g., Booch). The BTs goal was to adhere as closely as possible to the Annotated Reference Manual (ARM) definition. They wanted to translate platform-specific errors into platform-independent exceptions and provide built-in testability. Bedrock will allow you to see a log of wherever exceptions could be thrown and gives you the ability to identify what to handle. The error handler uses a significant subset of ARM and has automatic clean up of stack objects. Exceptions can be thrown from constructors. Exception types are classes which are organized into a type hierarchy. Bedrocks take advantage of the resource acquisition is initialization idiom. The exception hierarchy will have about 60 different classes that are defined into BEL (the Bedrock Exception Library). There will be sample code to demonstrate BEL.
Mike Burbridge discussed Bedrocks memory handling scheme. Mike said that the design goals and criteria was for: an average block size of 50-500 bytes, large request stream (10,000 blocks), 85% space utilization, time efficient, and transparent to the programmer. To reach their goals, they used a best fit allocation strategy, free-list stored in a binary tree, and made modifications for small blocks (<20 bytes) to reduce overhead per block. The good news is that theyve come up with a strategy that works fairly quickly (they showed test results to prove this).
Bedrock will have a platform independent resource language called BEDRC. This compiler will work similar to Rez and will produce a platform specific binary file. It supports views, menus, string lists, bitmaps, cursors, icons as well as user defined types. Data types can be inherited and can have recursive definitions. BEDRC has a full ANSI C preprocessor and will parse C++ header files so that it can go into classes and get constants. The compiler supports a mechanism for native resource definitions and can use the #pragma command. There is a resource editor under development.
According to Steve Levine, Bedrock product manager, they will have a developers release in May at WWDC. This will include demo code/technical information and may be limited to a select group. They will beta the product in Q393 and ship in Q493.
MacApp -> Bedrock
If it wasnt clear to you already, Apple and Symantec are not going to be able to provide a complete auto conversion for your code already using MacApp. To do that would have been as big a project as Bedrock itself. Instead, the Bedrock Team will provide tools to help. For example, there will be a tool to convert all view resources, but it wont convert custom resources. They will have a utility, called Witches Brew, which will analyze an applications source files and mix in MacApp 3 headers. The utility will create a text file containing MPW commands indicating the change that should be made in your code. You will still have to do the actual editing though. The team has taken the safe approach and will only be making changes automatically if there is absolutely no doubt in the translation (e.g., resources).
Last year, Apple showed off Newton a bit at CES. This time, they did it again. For those of you who are still skeptical, we saw them demonstrate a Newton. Yes, they used a stylus, wrote in script (not block printing) and it recognized the letters converting them to computer readable text. This thing is quite cool and it does work!
James Joaquin, Product Marketing Manager for Newton Core Software and Tools, led most of the talk. His first point was that Newton is not a small Macintosh, it is an entirely new metaphor. Newton aims to help people do what they do everyday - information capture, organize, and communicate. To capture, Newton allows users to enter in free-form text and graphics. Organization is covered by the address book, calendar and other applications. Newton communicates by allowing a user to print, fax, and transfer data to another Newton or to a desktop machine.
The Newton team is interested in developers producing 3rd party add-ons and software as well as having them as early adopters. There will be a wide range of tools over time and it will be an open platform where anyone can play. This is a new category with new applications and opportunities.
Newtons information architecture is designed for object-oriented data storage optimized for storing, finding and linking small pieces of information. This structure is so ingrained into the machine that developers will have to use it. The architecture is based on the frames technology - it is out of the AI world, but it is not a full blown knowledge system. Theyve included a number of standard data types to enable information transfer across applications.
For those who already use MacApp, youll be happy to know that Newtons hierarchical view system is similar to MacApps but built directly into the system. There are a lot of views already built in, ready to use. The architecture is quite rich and is extensible by developers.
To develop for Newton, initially, you will use a Macintosh. Eventually, there will be tools for developing directly on Newton. Expect good communication paths between PCs/Macs and Newton. Apple will be talking a lot more at WWDC 93 in May. In the interim, you can contact Newton Developer relations on AppleLink at NEWTON.DEVS.
Windows development and other talks
There were a few sessions on Windows development. Emergent Behavior spoke about porting your MacApp project to Windows. The main thrust here was to keep your user interface code separate so that it could be more easily rewritten.
Microsoft is continuing to try to make a splash in the Macintosh development arena. They are showing off products such as Object Linking and Embedding (OLE) and Microsoft Foundation Classes (MFC). They are trying to convince Macintosh developers to write for Windows, but to me, it still seems like the fox preaching to the hens in the hen house.
There were the typical talks on MacApp tips, C++, object oriented programming, etc Solution-Based Modeling (SBM) got a warm reception from the audience and, Apple spoke about Dylan some more.
Things have been kind of quiet in the Taligent camp. Theyve been not only working on the project, but also moving and getting more people on board. Taligent has been looking at long term trends in computing. Since the 1960s, systems people have been progressively moving the code common to applications into the OS, GUI, and into frameworks. Today, developing has become so complex that there are basically only enhancements to existing products and no new innovation.
The people at Taligent are trying to change the way development works. The idea is to bring the process back to the time that a small developer can build something. This is good not only for the small developer, but for larger ones as well. Their answer is to integrate tools into the OS and then let people approach it in a holistic manner. There arent many details - they are saying that they will ship something in the middle of this decade. Stay tuned
Shows like this always have a lot of products displayed. There is definitely not enough room here to do justice to these products, but here are a few that were shown: SmallTalkAgents - an excellent development environment for Macintosh SmallTalk developers (due out in the next couple months), Component Softwares Component Workshop, Serius Developer, BETA (now being published by MADA), ObjectModeler - a design tool, OOPC - a MacApp alternative, Camelot and Quorum - cross development solutions, SBM tools for more effective development solutions, Ad Lib - a ViewEdit replacement, VisualWorks - a SmallTalk Application Development Environment, NeoAccess - an OO database tool, IcePick - a MacApp View editor, First Priority - OO prioritization tools, POET - a cross platform OODBMS, and more
MADACON as a whole
There was a whole lot more that happened at MADACON - talks on managing large OO projects, how to port MacApp project to Windows, OOP directions - not to mention the excursion to SeaWorld and the reception following. All in all, it was a good solid conference showing off some interesting technologies and providing a platform for discussing the future of Mac development.
Neil Ticktin, Editor-in-Chief
The Publisher's Column
Hardware upgrades versus replacements
By David Williams, Publisher
So, last column, I asked questions about Apples new model-turnover strategy. No sooner had that coloumn gone to press, when Apple announced a whole new list of models, prices, and features. To make matters personally worse, MacWeek noted that a Quadra 800 has more bays and is actually slightly faster than my shiny new 950, which was marked down in price from when I bought it. Yet, the software that really makes these little boxes useful remains largely unchanged. Why is it that its acceptable for upgrade hardware to be brought out at full price, with a few exceptions, while new software is almost always priced very cheaply for upgrading? What does this difference mean?
To me, the main impact of this disparate pricing structure is that there isnt as much incentive to develop new software as new hardware, since new hardware generates new revenue while new software doesnt. The relative slowness of software development can probably be tied back to the fact that there isnt the big immediate payback of lots of full price sales.
I wonder if software producers couldnt learn a lesson here? What if developers brought out software designated comparably to hardware, so there would be, for example, QuickMail 950, which would be promptly replaced by QuickMail 800, at full price, of course. The hardware manufacturers dont seem to mind that their model numbers dont stay linear compared to performance, so maybe software producers shouldnt either.
The standard answer to all this is, of course, that hardware is more easily differentiated by the buying public than versions of software. So the hardware guys can get away with introducing a new model every six months while software guys cant. I wonder if this will still be true when the new models arent technologically much different than the old. The new Centris machines make justifying a Quadra very difficult for business purchasers. Yes, there are little differences that mean things to computer-types, but there arent the big differences that the average business buyer, like me, can really latch onto and parade before the board of directors.
Its an interesting question being posed by Apple, especially when upgradeability is the new watchword in PC design. Intel worked hard to be able to sell aftermarket customers an easy-to-install upgrade chip instead of a new box. It appears to me that the old lesson the software industry learned about upgrades versus replacements has now been learned by Intel, and forgotten by Apple.