The luxury of ignorance: A follow-up (NewsForge) [LWN.net]

The luxury of ignorance: A follow-up (NewsForge)

Posted Mar 2, 2004 9:33 UTC (Tue) by evgeny (subscriber, #774)
Parent article: The luxury of ignorance: A follow-up (NewsForge)

I wonder whether we're going to see soon the "Cathedral..." revised ;-).

To be honest, ESR should try hard to understand _why_ we're in such a loosy shape - having thousands of talented programmers, yet failing at some obvious places (wrongly thought-out GUIs for admin tasks is just one of many problems plaguing the FLOSS development; I can rant for hours regarding the pathetic absense of a general-purpose configuration API, for example). Maybe one should take less bias towards the bazaar model? Appearance of freedesktop.org and its emerging set of standards is IMHO a sign of such a maneuver, but we're in an ultimate need of lower, OS-level widely employed standards such as (the aforementioned) configuration, resource announcement/location,... APIs. The truly sad fact is such protocols AND imeplementations in many cases do exist, but are often used only by a handful of projects. Look at the SQL-related projects, for example. Among hundreds and hundreds of DB-based apps/utilities only a few per cent use the ODBC API. Granted, some are specially tailored for Postgres or MySQL (though why these two can't reach mutual compatibility while successifully emulating commercial closed-source counterparts has always seem to me a dark force conspiracy), but for most, the ODBC common denominator would be more than adequate. Etc etc...

(Log in to post comments)

The luxury of ignorance: A follow-up (NewsForge)

Posted Mar 2, 2004 15:22 UTC (Tue) by hereticmessiah (guest, #19909) [Link]

The first thing is to realise the problem exists. ESR has done this and done it in a
very public way, unlike the rest of us who knew it was happening but never really
crystallised it. Then the analysis begins. That's what we all have to do now.

The luxury of ignorance: A follow-up (NewsForge)

Posted Mar 2, 2004 23:57 UTC (Tue) by bojan (subscriber, #14302) [Link]

> The first thing is to realise the problem exists.

Yes, of course. But ESR was far from first to see this one. People that are developing GNOME, for instance, have written papers on the issue and keep discussing it publicly, and more imporatantly, are writing software the follows guidelines that were result of those discussions. So, ESR is simply reinventing the wheel here.

What he should have done is write patches that fix the problems he was experiencing and then write a piece that would celebrate open source development model. But he's telling us: "Do as I say, not as I do." I kind of remember him asking a lot of other people to "show us the code". Where's the code now?

His piece is just a rant. And at that, a rant directed at a completely wrong group of people, simply because he was too lazy/angry/misinformed to find out where the problem actually was. Simply put, nobody will be better off for his rant. The code will not be written faster and no unknown issue has been discovered. He just wanted a bit more attention.

The luxury of ignorance: A follow-up (NewsForge)

Posted Mar 2, 2004 16:24 UTC (Tue) by frazier (subscriber, #3060) [Link]

wrongly thought-out GUIs for admin tasks is just one of many problems plaguing the FLOSS development; I can rant for hours regarding the pathetic absense of a general-purpose configuration API, for example).

Yeah, I spent some time thinking about this problem recently. My incomplete thoughts are here under "Thoughts on GUI system administration, both locally and remotely":
http://www.userlinux.com/cgi-bin/wiki.pl?Brock_Frazier

"Configuration: A single point-of-contact for system configuration" is the most recent in there, and /ThinServerUI is out of date since it didn't take into account the "Configuration" library and is overly web-centric.

...and there's an overview of different UIs for the same tasks at different times here.

As I've mentioned elsewhere, I'd be very happy to work on the UI ends of this, but I don't recommend me writing (ugly) code for anything of this magnitude.

There's supposed to be something called debconf out there that tries to accomplish something along the lines of what Configuration would do.

All I know is I'm the sort of user who would gretly benefit from something like Configuration and the associated control GUIs but am not the guy to implement it, unfortuntely.

The luxury of ignorance: A follow-up (NewsForge)

Posted Mar 3, 2004 9:03 UTC (Wed) by evgeny (subscriber, #774) [Link]

I believe the real problem is not a GUI, but the principal lack of a unified way of storing/quering configuration parameters. Creating a layer _above_ the existing thousand and one configuration approaches/formats (like what Webmin is doing) is a dead end. We have to build a new, powerful configuration library (with C API, no dependencies except libc, and bindings to all reasonable scriptable languages including /bin/sh) and rewrite existing apps to use it. This seems a long way, but only this will do in the long run. Even a trivial .ini-style configuration API, if existed in libc, would have saved uncountable thousands of work hours - both of programmers and users. If there wouldn't be malloc() & friends in libc, each application would have to implement it. Thanks God, designers of libc thought of the memory allocation issues. But for 20+ years, the Unix world seems to fail realizing a unified approach to the configuration stuff is important as well.

The Gnome's gconf is something very close to what I'd like to see - EXCEPT its huge list of dependencies, rendering it useless for anything basic like system configuration.