Merlin Overview

Merlin is the name of a project started in 1984 to develop an advanced and integrated computer system, including programming language, operating system and hardware. Current plans include all this and adds versions that will run on standard PC hardware and even versions that will run on top of popular operating systems such as Windows 95 and Linux.

Many people would argue that the time for Merlin has passed, there is no longer any room for a new OS - specially since some of Merlin's unique technologies are starting to show up in traditional products. So the question is:

Why will people want to use Merlin?

Except for a few people who must always be different and have the latest and greatest, the answer is simply that they won't want to use Merlin. But they will want to run applications not available on any other OS.

Merlin doesn't exist yet and so has no applications. Established OSes have tens of thousands. Why would Merlin ever have applications that the others didn't?

First of all, while Merlin doesn't exist it already has applications since it will run programs developed for Sun's Self 4.0 (including text editors, a Web Browser, several simulators and many utilities). It will also be compatible with ANSI Smalltalk and will run programs developed for that standard.

It is important to note that application counts for popular OSes tend to be highly inflated. There tend to be a large number of abandoned programs that are no longer used even if they still work. Another large part of the applications are basicly design to fix or work around deficiencies in the OS and would simply make no sense in a new design (just think of the number of memory extenders that used to be popular for DOS - would a Linux user ever miss them?).

We can divide applications for a new OS into two groups: compelling apps (or killer apps, tractor apps or whatever you prefer to call them) which produce so great a benifit that it is worth buying a computer or an OS just to use them, and the "envy control apps" which take the pain out of abandoning the old OSes.

Envy Control

It isn't easy to adopt a new computer if it can't handle basic tasks like word processing or spreadsheets. One huge temptation for developers is to try to make these applications so good that they will actually be compelling apps. The problem is that platform envy is often more about compatibility than it is about features: it does little good to be able to claim that you have software much better than Office97 for your system if you can't read a Word document that someone sends you attached to an email message.

The initial solution for envy control in Merlin will be to be able to run Merlin, and its applications, on traditional OSes like Windows and Linux. This will allow access to all legacy applications without a heavy investment in making Merlin emulate other systems or be able to convert all data formats. As the number of Merlin applications increases, a second phase in envy control based on format convertion will make running Merlin on the bare hardware more and more attractive.

New and Compelling Applications

A new OS makes no sense at all unless it allows the development of really useful software that would be much harder to create on other systems. Unlike envy control apps, which have to be many to be effective, all it takes is a single compelling app to give an OS a fighting chance in the market. Having a compelling app "by luck" (someone happened to develop it using that OS) doesn't help for there will be a tremendous presure to port it to more popular systems and it will be done. A compelling app will only remain so if it depends on unique features not available on alternative OSes.

For developers, the reason to choose Merlin for the creation of a new program is to make money or to have fun (probably both!). It will be possible to make money with much less effort thanks to Merlin's use of superdistribution and it will take much less programming effort by the reuse of other people's objects and the elimination of these time consuming tasks:

• GUI Programming - Graphical User Interfaces made applications so much easier to use at the cost of making developing them much more complex. Luckily, window builder software have come to the rescue since the early 90s. The ability to manipulate working "morphs" with Self's object model (use can also see Morphic, a much simpler GUI framework than traditional ones, working in Squeak Smalltalk. But in a class based language it isn't as fun), however, takes GUI programming to a whole new level.
• File I/O - There are many data structures in memory, but disk files flatten everything out as a simple sequence of bytes. A large part of any application is code converting between the two worlds. Since Merlin doesn't use files but is based on persistent objects that are transparently stored on disk, none of this code ever had to be written by application programmers.
• Client/Server Communications - As more and more programs run on more than one machine, the implementation of communication protocols is becoming a significant part of development costs. Remote Proceedure Calls (RPC) standards greatly reduce this task by automating most of the work, but Merlin is entirely built on a "message passing at the bottom" model and a parallel programming model that can hide remote communication code inside the system once and for all.
• Cache Control - Even novices can write applications that work, but it takes experts to make them work well. Part of their secret is rewriting the sotware so that results are saved and reused instead of being repeatedly recalculated. The problem is that the transformed program is hard to understand and maintain. So Merlin will allow the programmer to add caches to the original program using annotations for the reflective layer of the OS rather than through code changes.

There will be objects to reuse simply because it will be easier to create them (see the above reasons) and it will actually be profitable to do so.

back to: