Forum Post
Index > Website > Nice work

Author/Date Nice work
29/08/2003 4:51pm
Hi Matthew Allan,

I read the news section and throught that I would post to answer a few questions and such...

I am glad that you are trying other Linux distributions than RedHat and Mandrake, Gentoo is probably the most BSD-like of the Linux systems. Debian is however still probably the best for long term use (the package maintainers are numerous and know their stuff since quite a while), and the package management (especially using apt) is great. It is actually the only Linux distribution which I still personally run.

It is worth noting that KDE or Gnome are to me only optional components, and easily replacable by the window manager of your choice. For instance, I found myself more often reverting to using simple icewm and gtk only applications (I'm a C coder, mostly in the networking server software field and find myself doing everything in a single X terminal (aterm) using the screen utility, and vim, and the only GUI applications I run are very rare)...

I saw in a comment that the Linux kernel seemed to be the best in your opinion, well such conclusions are usually very subjective... I personally prefer the clean design of the NetBSD system for instance (which even has abstracted the architecture-specific bus space to a layer)... This among other things helped make it the most portable free unix-like system around (and one of the most stable, too). The OS remains the same no matter what hardware you use, be it a PDA or a 64-bit alpha or ultrasparc. And it supports modern hardware such as gigabit ethernet, AGP and USB 2, wireless, iSCSI... and it's flexible licensing allows you to even make your own closed-source commercial system with it if you which. (MacOSX and Solaris/SunOs include alot of Mach and BSD code. Latest MS Windows releases, and QNX now use the strong BSD TCP/IP stack). Most BSD systems can also run binaries from a variety of other OS, including Linux native binaries, also. Most of my systems actually run NetBSD (but I keep a Debian Linux box for testing my software before releases).

As per the license bashing issues, I understand totally, you write your own software and are free to license it as you wish. I personally license most of my public stuff under a BSD-style license with advertizing clause, and of course I had to code commercial closed source applications (sometimes even exclusively for a corporation) to survive. To code, you have to also live :)

On a side note, your "Linux" version of your software would most probably work as-is recompiled on most other UNIX and unix-like systems)... If it was distributed in source unix release might be a better naming, actually. Of course that's not the case for binary releases.

So, just my ten cents :) Keep up the good work!
29/08/2003 7:54pm
Now that you mention it the *BSD's might make better kernels for my "frankenstein OS" but I'm not about to embark on that ginormous project just yet ;)

There is a standing offer that if any one ports LGI to a new OS then I will attempt to release builds of i.Scribe and InScribe for that platform (MacOSX included). As more than 50% of Scribe is in the LGI support library (which is LGPL), and the code in Scribe itself is actually 98% platform agnostic.

So far no one has taken up that offer. The rest of my programs are open source, so people can play with them. However noone has bothered to do that either.

As for my connection to KDE and Gnome, that is simple to intergrate some of the relevant settings. Mostly using KDE and Gnome's internal settings database for fonts settings, file -> MIME and MIME -> application mapping and support for putting an icon in the system tray applet. So far those two desktops have NOT made a compatible API for doing MIME stuff much to my disgust.

Otherwise I just use some sensible defaults for things like fonts and hope for the best.

My hope is that's stuff slowly becomes the standard for file -> MIME mapping at least. And that some 3rd party library becomes the standard for matching MIME types to applications, like fontconfig has being the standard for font configuration.

Thank you for your comments.
29/08/2003 11:29pm
The seems an interesting initiative indeed; The lack of such standards cause much more software than actually required to be installed on the average unix free desktop system.

Neverending dependancies among simple utilities are common case, and much less resources would be wasted if more definite standards were set. But, such is the way of the open source software also, everyone is free to have their alternative :) it makes it a little confusing.

As an example, yesterday I upgraded a utility called dia on a NetBSD box here, which now uses GTK2. Well, all the components of GTK2 were installed, and for some reason, python was also part of the dependencies... Then libglade2, with all it's own dependencies (which include gnome2 components)... Now what if I had a single KDE based application to install?

The result would be installing the kde3 libs, QT 3 and all their dependencies, which includes an ldap server, db3/4, alot of XML and printing related dependencies etc... Quite a mess, actually :) fortunately, pkgsrc simplified alot the building process, but it is still a little discouraging, and the reality of using a unix based solution with free software choices...

I think that I also saw notes about framebuffer based solutions rather than the standard X11R6. Well, X11R6 is great for networks where clients and servers can operate remotely; For instance several X11 terminals (running X servers) can run applications off a single server where the actual accounts and X client applications can be stored. However, I fully agree that XFree86 alone is much too slow for some applications, consider for instance accelerated GL, which could be far more optimized. And as you specified, an X11R6 compatibility layer could be provided over a better system...

Unfortunately, it appears that the best solutions will remain the commercial ones however, for a main reason: proper integration of a full system is costly in resources and management... There is also the need for some strong direction and sencorship to unite the project goals and limit code duplication, as well as control quality. Moreover, commercial support for the product is also required by hardware vendors (I.E. OpenGL drivers for the NVIDIA need to talk through a specific and defined API), they'll use that API (and/or ABI) if it is known to be widely supported only, like you will port your software to the most common platforms only. Ah I btw now also remember that stronger sencorship is probably what makes projects like NetBSD more interesting than Linux to me (more time is spent working on actual solutions to avoid temporary hacks, which usually get rejected rather than accepted to bloat the existing code and compete with other similar half-working solutions. There is no such work on it done on the GUI level however, still, as it never was considered as an eventual ideal desktop system by the project core members).

I'm not used much to HTTP forums, it's been a little adventure out of the common on both the forum and topic, actually. It's been nice exchanging.