Elektra use flag

[chuliomartinez]

Hi, I been watching this project for some time now. They developed a lot of patches so I wondered if somebody would be interested in adding a "elektrify" use flag, thus compiling the app/lib with electra support. It is not easy to try a project like this by hand as there are to many things to be converted etc, but gentoo seems to be the top choice because of the USE flags power.


http://elektra.sourceforge.net/
_________________
naked&dead

[nevynxxx]

Your best bet would be to try it yourself.

Look at the ebuild docs on the gentoo website, look at some ebuilds that patch as they build, then add the things you want to the ebuilds you want as you go.

You can then submit your new ebuilds as a feature request in the bugzilla when they work....
_________________
My Public Key

Wanted: Instructor in the art of Bowyery

[c07]

https://bugs.gentoo.org/show_bug.cgi?id=63417

[Flamie]

This is a great idea, and I suggest you also draft a GLEP.
gentoo is a perfectly good testbed for incorporating elektra, because of how it ends up as the users choice, and will probably be turned off by default.

I would really want to see this ebuild and patches making it into portage, as well as I want to see gentoo promoting the use of elektra for new apps and patches on old.

again, great thing, lets just hope a big number of people understands this :)

[rauar]

[binary_runner]

I've read threads, I've read project page. First time I thought it's good idea. Yes, the idea to have some unified form to access configuration options is good, but the solution is not.

From my point of view elektra is just a plain demonstration of messiah syndrom - I don't like the design and I'm sceptic about implementation.
Let's look at ti as a software project -- I think all-in-one sollutions are aout of fashion for quite a time, its much better to solve the core problem and not to close doors to associated issues and to solve it as transparently as possible (no external patches, they are always late).

So what's the core problem IMO ?

* location of config files
NO. You have memory (your's or community's) and tools like find or grep. Once you know, where the hell the config file is, you can easily navigate there with shell tab completition and inside the file with your editor searching. Plus with some app specifig config files you can tailor the structure and fragmentation of configuration subsections to your needs. note: I'm happy to say gentoo has very good /etc structure -- as little levels as possible and almost everything is easy to find thanks to simple and well-commneted config files or scritps.

* format
COULD BE. But the key-value pair "structure" is dumb as a rock. I don't want pile of incredible long keys pointing to some strange values, it gives me no benefit, just head ache. I don't need data, anybody does not need them, we have plenty of data, too much to be honest,. What we need is structure corresponding to meaning of the things -- let's call it semantics. If I could vote for one and only one format, I'd use XML -- it can be a bit wordy it is about structure, has a lot of tools and it;s here for some time (so most of problems are aither solved or somebody works on them. XML has tree -like structure (yes I know, the elektra can be viewed so, but not with text editor or for the price of additional layer), comments (don't see any in electra), namespaces, ....

* stability
IMPORTANT BUT NOT CORE. We have plenty of choices to make system configuration more stable. Adding one library will notprobably make things worse but certainly not better. As for stability, we need validation, transcations and backups.

* access control
I WOULD SAY YES. I would like to allow users to modify parts of configuration files and have well-formness, schema and user specific validation before they are saved to disk and cripple the config file. We could do it for applications wit XML config files by creating XML filesystem presenting certain parts of document (whole branches or values) as ordinary files and calling external applications for schema or user-related validation. We don't need modify current appliation we can just wrap them or try convince them to XML version of their syntax. System configuration can stay as it is and probably later move to this approach too, if it proves to be usable.


Summary : Elektra is bad solution for real problem(s). It will only add another incompatible piece.

Now tell me where I'm wrong and convince me how elektra is cool.

[jsosic]

XML is very bad for configuration, and it's even worse idea than Linux implementation of Windows Registry.

Idea behind elektra has some strong points, like single lib for configurations, various backends, and so on. But, it's flaws outnubmer it's good points Look what a mess registry did to Windows... And they've started to abbondon the idea. I don't know why is Linux following Windows so blindly, and makin' the same freeking mistakes... I really don't get it. An insane man does the same thing over and over again waiting it to produce different outcome - but IT WON'T !!!

In my opinion, elektra problem is problem of attitude. If project like Apache, /sbin/init & Samba drop their respective configurations, and force sysadmins to use elektra and it's tools, than it's really really bad. Linux is becoming more of a bad copy of Windows every day, and this Registry idea is another small step in that direction, but when/if we start to loose freedom of choice, and when things start to be forced upon us, then all this GNU/Linux stuff will loose all meaning. Why not just buy the original (Windows), and forget about some cheap copy(Linux)?

I'm really really sad, because I have a hunch this won't be an option, and it won't be negotiable
And the worst thing, there will be no FreeBSD to run to cryin', 'cause elektra will be there too
_________________
I avenge with darkness, the blood is the life
The Order of the Dragon, I feed on human life