For compatibility with the Object Pascal memory model, which ameliorated the lack of physical memory management, C++ on the Mac generally used handle-based heap objects. This had serious structural issues because they just weren't firmly located in memory.

Ultimately regular pointer-based heap objects won out even on Mac and reference counting became usable. I encapsulated the necessary activities in the smart pointer ZRef and the object base class ZCounted, which stabilized in the late 1990s after some struggles with the advent of preemptive multithreading.

All of which is to justify why I haven't simply embraced the C++11 standard library's smart_ptr — by implementing reference counting for myself I think I stumbled over every possible mistake, but also found a really useful mechanism I haven't encountered elsewhere. In the literature reference counting is considered to have two fundamental problems — reference cycles and object resurrection. In my experience a cycle (ultimately) indicates a broken data model.

However, inadvertent resurrection is a significant problem.

The issue is that we may want to take additional action when an entity's count transitions from one to zero, but if that action itself needs to treat the entity as counted, then the count will necessarily increment from zero, must ultimately transition down again, and we're into recursive execution of finalization.

Another situation is when we don't want simply to release the entity when it's no longer in general circulation, but want to change its treatment, perhaps by placing it in a cache. So we actually still have a physical reference to the entity, but managed with some different mechanism. This kind of changeover is very susceptible to thread issues. We could acquire a lock every time we manipulate any reference count, but that utterly kills performance and creates a pervasive and non-obvious source of deadlocks.

The solution is in the statement — we know to take action when the reference count changes from one to zero. What if we don't change the count in that case? Remember we're tracking the number of extant physical references. Seeing the one to zero transition tells us there's no other reference to the object than the one in our hands at that moment. So any further reference originates from the one we're considering finalizing. Rather than the typical atomic-decrement-and-test, releasing a reference uses a get, test and CAS sequence. The finalization code is now free to pass the entity to other code and no (legal) operation can again take the count from one to zero. Ultimately execution returns to the finalization. If the count has returned to (or remained) one, then we release the entity, otherwise do a simple atomic decrement and exit.

Transfers (by subscribers) use perfect forward secrecy, and can be entirely asynchronous, so sender and receiver need never be online at the same time. Of course if they are, then transmission goes as fast as the physical network allows.

We built the backend and client mainly in Python, supporting Mac and Windows back to 10.5 and XP respectively. For Linux we support 32 and 64 bit Ubuntu 12 and later.

Much of my focus has been on keeping OS variations from impacting the main body of the code. To that end, we're using wxWidgets to abstract the GUI, and I've implemented such native code as needed to interface with unwrapped aspects of the OS and third-party libraries. In some cases the fluent Python bindings to OS facilities just aren't fast enough, and native code is used there also.

Graphic and game design by Wendy Carmical, project management and game design by Maura Sparks.

It's written in C++ using OpenGL, and thus is completely portable between iOS, Android and OSX.

In this enhanced version of SPARKvue sensors are tagged with fiduciary markers. The video from a camera pointed at the experimental setup is shown beneath normal UI elements, and measurements from sensors are drawn into the video as it runs.

In ZooLib I made several attempts at creating a clean suite of cartesian types and operations, with conversion operators and constructors, pseudo constructors and miscellaneous helper functions. But the approach just never scaled well.

Recently I've taken a very different tack. In ZCartesian I've defined a suite of template accessor functions -- passed a rectangle/point the edges, corners and centers can be extracted. Offsetting and aligning operations build on those accessors, as do sensible algebraic operators.

The accessor functions themselves forward to functions defined in traits templates. There's a suite of base classes for rectangles that are defined by variations of origin/extent or left/top/right/bottom. Supporting a new suite of cartesian types can be as simple as providing the definition of a traits class inheriting from one of the provided bases.

In this way we can work with NSRect, CGRect, XRectangle, QuickDraw Rect, Windows Rect, SDL_Rect, and their associated point (and size) structures using identical notation. And when we need to switch representations there's just a single pseudo-constructor to do the job.

Generally a proxy is a long-running system service, and it builds and evolves knowledge of what outbound paths and upstream proxies are available and usable. This proxy launches when the device connects, and shuts down when it quits, so any knowledge it can acquire is likely to be stale.

The approach I took was to try everything at once -- initiate connections on every combination of network interface, proxy name and resolved IP address, and use whichever wins the race. Thereafter it starts with the last thing that worked, and only scatterguns after a timeout. Interestingly this makes regular desktop browsing more responsive too, but it's pretty abusive to the network infrastructure and is only safe for idempotent GETs.

The other interesting feature was handling Proxy Auto Config (PAC) files. These are chunks of javascript that take a desired host name and return the name/address of the proxy that should handle the request. I'd never come across one in the wild, but they're very common in corporate networks where there's free rein inside the firewall, but all external access is via proxies. The system facilities on OS X are sufficient that the task can be handled by calling CFNetworkExecuteProxyAutoConfigurationURL. Windows on the other hand can almost do the job. The WinHTTP library can handle a PAC specified with a network URL, but not with a local file. So when we see that the PAC is a local file we fabricate an HTTP URL referencing a port we're handling, and any request on that port returns the content of the PAC file.

Zorap ran as a browser extension, so could be instantiated and destroyed many times within the life of a process. Any code it used had to be robust, and could not depend on operating system facilities to dispose of resources like network sockets.

Most computer languages provide a suite of atomic data types, values that are generally not themselves subdivided. These usually include numbers, booleans and strings, may also include raw binary, and may make distinctions amongst various sizes of floating point and integer numbers.
Languages generally include an aggregate type that is called an array, vector or list, being a sequence of values. Languages also usually include a hash, dictionary or map type that is a collection of name/value pairs indexable by name.
Persistent data structures and streamed protocols very often reflect these divisions -- a larger or smaller suite of atomic types plus the two aggregates. If you squint you can see the commonalities amongst the QuickTime/MPEG4, bencode, plist and json formats. Various flavors of XML-based serialization fit the pattern, and even the venerable CSV can be treated as a sequence of maps of atoms.
YadTranscode reads sources of several formats: bencode, json, quicktime, tab/CR tables, xml plists, a couple of stylized xml formats and a custom binary format. It transcodes that data and emits json, xmlplist, a type-rich text format akin to Next-style plists and a custom binary format.
It's built on some of the wide suite of ZYadR entities in ZooLib.

BBDaemon runs as a user-space process, not a kernel extension. The installer simply puts the executable in /Library/PrivilegedHelperTools and a launchd control file in /Library/LaunchDaemons.

RIM's own Mac applications (PocketMac for BlackBerry and BlackBerry Media Sync) take a different approach. They each install a kernel extension which grabs any USB-connected BlackBerry, and prevent any other application (including each other) from being able to talk to it.

The installer we've posted was built with revision 533 of ZooLib. You can check out the source and build it yourself:

svn export -r 533 https://zoolib.svn.sourceforge.net/svnroot/zoolib/trunk/zoolib

svn export -r 533 https://zoolib.svn.sourceforge.net/svnroot/zoolib/trunk/zoolib_samples

cd zoolib_samples/BlackBerry/BBDaemon/osxpackage

sh build.sh

The executable and launchd file must be owned by root/wheel, so the build.sh script uses sudo to change the ownership – thus you must be logged in with an administrator account and you may be prompted to enter your password.

ZBlackBerry can also make a BlackBerry connected to a PC accessible via the same API as is used when it's connected directly to a Mac. This lets Mac application developers run their BlackBerry application under the Windows-based debugger and still have it communicate with their Mac application. This is a crucial capability when writing anything more than the most trivial application.

ZBlackBerry is part of ZooLib, our open source C++ library.

The BlackBerry application is written in Java. It maintains a single communications channel with the Mac application, over which it receives requests to make, use and break network connections. The channel is multiplexed with ZooLib's Java ZStreamMUX on the BlackBerry side, and its C++ ZStreamMUX on the Mac.

Much of the Mac application's functionality is a C++ HTTP and port-forwarding proxy server, most of which is in ZHTTP. Its UI is implemented in Objective C using Cocoa Bindings.

The remainder of the Mac implementation effort was in reverse-engineering the BlackBerry/Host USB communications protocol.

Frankly, WebDAV is an inelegant protocol. It puts some information in the URL, some in header lines, and some is encoded as XML in the body of requests and responses. But it has three sterling characteristics:

• It exists
• It works
• It's usable out of the box on Mac, Windows and many flavors of UNIX

Turning an application into a service is straightforward, but involves a certain amount of boiler-plate code, and requires that the application be structured to be controllable by a thread started by the Service framework. The ZooLib class ZWinService implements the boiler-plate, and at its simplest requires only that a Run and Stop method be implemented on a derivation of ZWinService in order to connect an application to the service framework.

There are two problems with connecting a UI to a transaction-based store like the tuplebase. First, how do you efficiently identify the minimal set of interesting changes that have been made in the tuplebase. Second, in reading or writing data, what do you do when a transaction fails to commit?

ZTSoup addresses these problems by maintaining objects locally that record a snapshot of the subset of the tuplebase in which the software is interested. Local changes are recorded in those objects, and they are sent to the tuplebase as a single atomic update operation. It's simply a matter of hooking a call to the soup's Update method into the normal event loop -- all additions and removals of interest in croutons and sieves are sent to the tuplebase, and remote changes to the croutons and sieves are returned.

Because the soup may be talking to a tuplebase on the end of a comms link the update work is split into a synchronous piece, which executes in a small amount of time with no blocking, and an async piece that accumulates work done by the synchronous piece and communicates with the server as fast as latency and server CPU will allow.

But native code can't be included with unsigned applets, so I ported ZTBRep_Client from C++ to Java. ZTBRep_Client is the most useful tuplebase implementation in that it talks to a ZTBServer instance on the end of a comms link, generally a network socket. It's also fairly large, and the async nature of the protocol makes things somewhat tricky. The normal Java synchronization model is based around monitors, which are not directly useful for implementing certain types of concurrent processing, so I implemented condition vars and mutexes to provide more appropriate tools.

New features included support for upload and download of files as attachments to views and notes, headings within views, full text and attribute-based searches, and notification by email. It also supported graphs of useful metrics, implemented across platforms using ZDC_ZooLib.

On the back end, a significant change was moving from the original blockstore implementation to one based on B+Trees that uses a neat upward propagting copy-on-write of nodes to make it virtually bullet-proof - only a physical failure of the backing store can cause data loss.

Language agnosticism is a key feature of tuplebase. Another is its runtime flexibility, which contrasts with most relational database implementations.

The power of a relational database is that the data it stores conforms to a rigorously defined structure. For a single piece of software it's always possible to find a single structure that supports all the work to be performed. As the nature of that work changes the database schema is evolved along with it. But this evolution must be coordinated so that all software that depends on the schema is updated to accommodate the changes, or virtual views are created to preserve a logical interface to the data.

Tuplebase takes a very different approach. Rather than enforcing a structure it allows any structure at all to be stored. Different pieces of software with overlapping concerns obviously must coordinate amongst themselves, but there's no requirement that there be any single authority managing everything. And because there's no mandated structure it's possible at runtime to treat multiple physical tuplebases as a single logical tuplebase, even if they contain data of different provenances. If the code using the data is written to be accommodating to the gradual migration of meaning then newer code can correctly interpret older data, and older code will ignore (but preserve) newer data.

For a time we had native client and web access running simultaneously, but never in a released version. The 3.x client remains buildable as of May 2006, but still talks only to a 3.x server.

The ZooLib file API doesn't address more esoteric file systems like VMS or IBM's MVS. It models a hierarchical system, with accomodations for special roots (like Windows' UNC paths and Mac Classic's multiple volumes with the same name). Files are byte-oriented, and as with other stream operations ZooLib's file Open and Create return a ref counted streamer object, and optionally an error enum providing more detail in the case of failure. The representation of a pointer to a node in a file system also follows ZooLib's practice of a using a smart object with value semantics.

The ZFile suite was the first place I formalized the notion of a 'trail'. This was another case where it was important to come up with a new name that was suggestive of an existing notion, but distinct enough from it that its different semantics could be defined. Picture a tree; a trail is simply the list of steps to be taken to navigate from some node to some other node. A step is either to traverse the link from the node to its parent, or to traverse the link from the node to one of its children. We represent a step as a string. An empty string indicates a 'bounce', a traversal to the parent of the node. Any other string indicates a traversal to the node with that name. Whether the name is attached to the child node, or attached to the link to the child node doesn't matter in this model.

The notion of a trail through a tree shows up in assets and in the handling of HTTP requests. Theoretically it could also be applied to tuples but I'm not quite ready to harmonize everything to that extent.

So, a ZBlockStore takes a single file (or file-like entity) and allows an application to work with resizable blocks allocated within it, easily, efficiently and (with ZBlockStore_PhaseTree) safely.

• Easy to use — blocks are managed with the same API as real files.
• Efficient — all the data lives in a wide fan-out B+ Tree.
• Safe — modifications to nodes in the tree cause the entire ancestor branch to be replicated into existing free space, and the new root is atomically updated only when descendant nodes are known to be in stable storage.

ZooLib abstracts threads, files, networking, graphics and the user interface. And there are escape hatches whereby OS-specific features can be accessed without compromising the platform-agnostic nature of the rest of the code.

ZooLib has been under continual development since 1992, and incorporates fixes for every OS bug or anomaly I've encountered, and generally provides cleaner mechanisms for most tasks than OSes traditionally provide. It's been the basis of Measurement in Motion, NetPhone, and every version of the Knowledge Forum server and client applications.

By this time ZooLib abstracted Mac OS, Windows and UNIX, so the server could be hosted on all three OSes.
The text editing in Knowledge Forum 3.0's note windows used the PAIGE text engine, which was a fairly crusty 60,000 lines by this time, but was portable across Mac and Windows. So although the rest of the client could be built for UNIX, without the text editor it wasn't particularly useful and we never released a UNIX client, although that wasn't a big deal in 2000.
Unfortunately we also had to abandon Kevin Parichan's excellent painting implementation, as it was also Mac-specific. Instead we used ZooLib's NPainter, which used only ZooLib features and was thus completely portable.

A key design feature is that it never touches its hosting environment directly. Instead it simply posts update notifications, and so the hosting environment is free to clip or draw on top of a painting as it is being worked on. This makes it easier to host in environments with tricky constraints, and to achieve effects like multiple views onto different parts of the same painting.

Whereas NetPhone allocated some space for a stack to be used when invoking a stack-hungry function (the GSM codec generally), ZFiber calls setjmp to record the current stack and registers, then calls longjmp on the setjmp buffer previously initialized by the next fiber to be executed. Because we have complete control over the entire process this makes it very easy to turn async callback-based functions (like most of the Mac's IO subsystem) into blocking function calls. And the whole thing can run at deferred task time, simply by having the callback for an async function schedule a deferred task which calls longjmp to the first fiber that's marked as being both runnable and safe to be run at deferred task time. A fiber can change its deferred task time eligibility on the fly, so sections handling IO can be run at deferred task time, and thus not be subject to the latency imposed by having to wait till GetNextEvent, but sections that have to touch non interrupt-safe sections of the ToolBox can be run exclusively at normal time.

CSILE was pioneering and innovative, but in May 1994 limitations in its feature set and a desire to see it in wider use led to a collaboration with Learning in Motion to redesign it. The scope of the necessary changes neccessitated a complete reimplementation and the decision was taken that it would be released as a commercial product.

Cyrus Ghalambor started work on the KF 2.0 client application in July 1995. I was busy with NetPhone till December 1995, when I started working full time on the client's data API and the server that would satisfy its requests. We looked at several existing database libraries, but they were either too limiting or could not run on Mac OS (Classic, of course). So I implemented ZDBase, a portable database engine that was built on ZooLib. Client and server communications and the server's administration UI were also built using ZooLib.

The client application's event loop, menus and windows were handled by a refinement of Berkeley Systems' Zoom framework. It had two main types of windows. The note window used Zoom for its controls, and the WASTE text engine for its text editing panel. The view window provided a two-dimensional icon-based representation of collections of content, its operation was closely modeled on that of Apple's Finder, and it was implemented using ZooLib. The note and view windows also contained pictures, in-line with the text in the note window and in a layer behind the icons in the view window. Pictures were edited in place using a nice painting implementation by Kevin Parichan.

Other dialogs and property inspectors were based on Zoom or ZooLib, depending on whether they were written by Cyrus or by me.

This was my first serious use of B-Trees, the essential basis of most high-performance disk-based storage systems. The searching of B-Tree content and the insertion of new data is well documented in numerous sources, but the deletion of data is invariably left as an exercise for the reader, which was a bit frustrating but ultimately empowering.

It was the commercialization of a project from the BBC's Interactive Television Unit called 'Future Worlds'. Future Worlds was investigating how to create video content that could be used in a traditional linear documentary and also in a hyper-linked computer-based environment. I worked at the BBC for all of three months before the ITU was spun off into an independent company, the MultiMedia Corporation. MediaMaker was published by MacroMind, and I went along with it to San Francisco.