Current mood:	sleepy
Entry tags:	code, linux, rant, tech

The changing face of Linux.Ars
So the past several weeks have seen a cooling of interest in Linux.Ars from many #linux folks; in particular, Jorge, whose baby Linux.Ars was, has been turned off it by some insensitive comments in the Battlefront. Eric Bangeman is taking over the management of the column, and intends to set up a sort of network of contributing editors and writers, assigning the responsibility of organising and editing Linux.Ars to whomever has the time and comes forward to do it.

I have a draft up of a writeup on internationalisation and localisation; let's see if it makes it in Wednesday's issue.

As far as the Mono patent bandwagon is concerned, I'm not worried nearly as much about patent infringement/enforcement as I am about performance. It just strikes me as a bad idea to be moving in the direction of requiring nearly every piece of GNOME software to be in the form of bytecode (Intermediate Language, whatever). I can only anticipate software becoming more CPU- and memory-hungry and less responsive, as it needs to be binary-translated every time it runs (and even if the binary translation is stored, it may never be quite as efficient or fast as native-compiled code).

What's wrong with a properly-designed object model based on a (perhaps slightly stripped-down) CORBA middleware/IPC layer with services published and subscribed to? If the object model is done properly enough, with good-quality language bindings, we've realised the dream of a language-independent runtime, but without the crummy bytecode interpretation/dynamic binary translation.

As far as I can see, the primary advantage of the .NET framework is its set of class libraries. What is to prevent such libraries being built for a regular middleware-based runtime? Sure, the runtime might not allow for complex class hierarchies, but what is to prevent a middleware-based 'wrapper' around such an implementation, with the language bindings tidying things up? We don't need no stinkin' common-language runtime. (I guess I'm advocating a somewhat cleaner alternative to Microsoft's ATL, except it'd only be a wrapper around the middleware proxy, where the actual type handling/object instantiation/etc. would be done at the runtime end.)

The .NET framework is presented as a one-size-fits-all solution, but primarily oriented toward distributed/multi-tier applications. It's overkill for a desktop environment. It might sound nice and buzzword-compliant, but I have to question the practicality of Miguel de Icaza's intent.