The Northern Spy
Prognostications and Reviews 2007
Last year's Northern Spy prognistications
turned out about as one might expect, given their pedestrian nature and the relative predictability of the market these days. The Intel switch went faster than most people expected, beginning with the portables as we expected. New iPod models, and software versions were also served per the menu, though the Spy's forecast of designer colours for portables stopped at black and white. Perhaps this was only a first step and there's more to come, but if so, back-to-school or Christmas markets are the ones to target, so we won't see apple green cases for a while.
Leopard is on the immediate horizon, and Apple has apparently taken delivery of quad core chips. It would make sense to introduce a high end desktop and/or server simultaneously with the new OS cat. A dual chip quad, that is an eight-way tower, would be an easy prediction for the near future. A nifty suite of applications to sweeten the pot would also go a long way toward selling to a whole new group of buyers. Face it. POVNM (plain old vonNeumann machines) are out.
The reason manufacturers are going with multi-core and multiple CPUs is that chip makers have hit the wall on speed with current and foreseeable technologies, with 4GHz apparently the glass ceiling for now (heat, quantum effects). Thus a multi-core approach is the only way to increase total system throughput. Under the hood, the OS (and some programs) factor their tasks and farm them out to the various processors to work simultaneously, then reassemble the pieces of the solution afterwards.
However, all this has limits. First, not every program can make use of multiple processors, either because it has not been coded to do so, or because the nature of the problem is not suited to factoring processing this way. Second, the real bottleneck in many, if not most, applications, is not processing speed but I/O. Since there are already separate I/O coprocessors, adding more CPUs can scarcely relieve this aspect of traffic congestion. Third, the overhead involved in doing the factoring and reassembling increases with the number of processors and for some types of problems is an appreciable fraction of the actual processing time, so that beyond a certain point, adding more CPUs slows the system down. Thus, barring radical changes in chip technology, we cannot expect that throughput will continue to improve at historic rates, and indeed, there may be overall system limitations in view. Moore's law may be suspended or repealed.
With respect to market share, Apple still appears on track to reach the eight to nine percent range by the end of 2007 and the ten percent figure as early as July 2008. However, any new product breakthrough (iPhone, iReader, iSonOfNewton) or any new MS stumble (yet another Vista delay, more security problems, continuing employee morale declines) could dramatically accelerate Apple's gains. Moreover, the market reaction to Leopard vs Vista in a new hardware environment could swing things Apple's way too. Indeed, it is difficult to imagine a scenario favouring MS over the next few years.
Oh, and a couple more things. 2007 is likely to be the year of Blu-Ray and e-ink, thus possibly heralding a return to prominence for Sony. Expect to see both in Apple products sooner rather than later.
Software noted in passing
include version 5.7.1 of Drag Thing (reviewed here last month in version 5.6) and available from the manufacturer or Northern Spy's download site as a free upgrade. This version introduces dock themes and fixes Intel and application compatibilities. Still recommented.
Another venerable utility recently upgraded is PopChar from Macility, now at version 3.1. PopChar has been rewritten from the ground up, is now universal, and sports an improved user interface, with easier access to characters and fonts. The Spy cannot remember a time when he hasn't had this program installed, though he won't offer it on his site for download because it isn't shareware but commercial. Still, it's well worth the $29.99 (online price--includes two years' updates) because when you need a special character in a document, you don't want to have to look it up. Just move the cursor into the menu bar, pull down the PopChar menu, and select the character you want typed into the front window.
Of less, er, utility:
The strength of small programs is that they do their chosen tasks. All-stars, like PopChar, Graphics Converter, and DragThing don't try to be Swiss Army knives. They deploy a single blade to do what they do very well indeed. Would this were so for all specialized utilities.
LogoDesign Studio from MacWare purports to be just what it's called--a specialized graphics utility for designing logos. The Spy picked up a copy to design new ones for some of his web sites. Not so easy.
To start with the preferences are greyed out. Is this a bug, or are there no program preferences? When making a shape, even a triangle, an attempt to have a border more than one pixel results in no border at all. Changing the properties in such a case does not take effect until the cursor is moved over the palette area, whereas, when changing the border thickness with a rectangle, the change is instant. There is no copy and paste, and there is no undo! If you cancel a save, the work sometimes vanishes (as in the window goes away without the work being saved). The use of the text tool to add taglines is completely nonintuitive. If you select the text tool, you have to click on the work area, then resize a text box to create a text object, but the resizing function is stuck "on" and no combination of keys will stop it. The "right" way (found by accident) is to select an area with the text tool active. Then the text box handles stay put. There is no way to guess this behaviour. The printed document is a slender book heat tells the user little. The online documentation is extensive, but alas, suffices itself by saying how to do some things, but not giving any examples to determine what the instructions mean. For instance, It's not clear what the Marquee is for. Bottom line: give this one a pass until it becomes a finished product. As it, it's unuseable.
New books this month
and, as the last time we reviewed programming books, both are from O'Reilly.
The second O'Reilly offering this month is Java Generics and Collections by Maurice Naftalin and Philip Wadler. This book covers the most important addition to the Java notation since it was first devised.
Generics have roots in ML and Haskall, and in more modern times, in C++ templates, Ada generics, and Modula-3. During his time with the ISO SC22WG13 standards committee , the Spy devised the Modula-2 generic library module and refining modules, then wrote the standard for both, so he knows how useful this paradigm can be.
Strictly speaking, generics aren't necessary if a language has OO. However, they are very useful in creating parameterized data structures and algorithms. Why write the code for a queue of reals, then repeat all that work for every other data structure you also want to queue? Why write a complete quicksort for every type of date? If a language supports generics, one set of code, suitably parameterized, handles all such situations.
Naftalin and Wadler do a good job of explaining how this all works in the recent addition to Java, correctly noting that although Java generics looks like C++ templates, the semantics are quite different (more lightweight) as only one set of code is generated, not one for each refinement. Personally, the Spy thinks of Java as Smallttalk dressed up to look superficuiially like C++. They authors also run the reader through the supplied collections. (These play the same role as the STL in C++, but appear to be cleaner and simpler.)
Another good book, though only time will tell how robust these collections libraries are. Its a pity the Java language itselfhas suffered so much under the stewardship of Sun (inventor of the notation). Had Sun chosen to cooperate with (instead of stiff) the standards community years ago, their notation might have gained greater acceptance and become more truly cross-platform. Perhaps Java generics a nd going to open source as they have now done will help on both counts, but the Spy suspects it might be a case of too little and too late. If you are a big time Java user, however, buy this book. Generics are the right solution to the two classes of problems it tackles. Would that more language designers knew this.
The Lord bless y'all in the new year.
--The Northern Spy
Rick Sutcliffe, (a.k.a. The Northern Spy) is professor of Computing Science and Mathematics at Trinity Western University. He's written two textbooks and several novels, one named best ePublished SF novel for 2003. His columns have appeared in numerous magazines and newspapers, and he's a regular speaker at churches, schools, academic meetings, and conferences. He and his wife Joyce have lived in the Aldergrove/Bradner area of BC since 1972.
Want to discuss this and other Northern Spy columns? Surf on over to ArjayBB.com. Participate and you could win free web hosting from the WebNameHost.net subsidiary of Arjay Web Services. Rick Sutcliffe's fiction can be purchased in various eBook formats from Fictionwise, and in dead tree form from Bowker's Booksurge.
The Northern Spy Home Page: http://www.TheNorthernSpy.com
The Spy's Laws collected: http://www.thenorthernspy.com/spyslaws.htm
The Spy's Shareware download site: http://downloads.thenorthernspy.com/
WebNameHost : http://www.WebNameHost.net
WebNameSource : http://www.WebNameSource.net
nameman : http://nameman.net
opundo : http://opundo.com
Sheaves Christian Resources : http://sheaves.org
Arjay Books: http://www.ArjayBooks.com