Slashdot Mirror
Online Repository for Hardware Configurations?
Great_Jehovah asks: "I've done a lot of trial-and-error and spent time researching getting various devices (e.g. motherboards, USB devices, DVD burners) working on Linux. I've also spent a bunch of time configuring different pieces of software for particular applications. I would like a nice centralized place to share these pieces of knowledge and also to see what others have done. I've looked on Google but either I can't conjure the right keywords, or this place just doesn't exist yet. Anyone know where such a site exists? If not, I'll start one."
No, no, no. This is a GREAT idea - think about it like "C Code Snippets" for Linux configuration files. Here are a few examples:
Case #1
I need a config file for Samba to emulate an NT4 PDC. I download the prewritten boilerplate config file, change the domain name, put it in
Case #2
Or how about a config file for Sendmail that uses Spam Assassin for spam filtering and renattach for viruses.
Case #3
I want to turn my old 386 into a Linksys-style NAT box. The only imbound port I need mapped is my web server (port 80). You got config files for that?
Case #4
Shared email address books with LDAP. I want to run an LDAP server with slapd to provide shared email address books, but I don't want to use LDAP for any sort of network authentication. I just want users to be able to create folders and contacts and move the contacts around in the folders (and add, change, delete and update them, of course.) Apparently, I'm the only person that ever thought of doing this because I haven't found ANY docs anywhere that describe this sort of thing.
Your target audience here is beginners and administrators who are migrating to Linux services who want to get things working without having to read and decode **ALL** of the documentation up front.
Let's face it, with if you take any given piece of software,there really are only a few different basic configs most people want to start out with, and once you get basic functionality in place, you can tweak to customize from there. Hey, they did it with sendmail, right?
"Lawyers are for sucks."
- Doug McKenzie
That Linux hardware database, IIRC, was hosted by a company that might have slipped under the waves with the fizzling of the dotcom boom.
But you have a really great idea.
Newbies and veterans alike would find it useful to know if some piece of hardware would work under a particular version of Linux.
Old hardware is important in cases where there isn't a lot of money to keep up with the latest hardware releases; schools, charities, and even businesses in the undeveloped world would benefit from such a knowledge base.
Also, if there were some means of making a spam-free 2-way communications channel from the site hosting the hardware db, it might be useful to kernel developers who want to know if their patch might cause a bad interaction in some corner case of two or three unusual old pieces of hardware that they don't have access to. They could send email to the owner of the machine with the configuration they want to test against asking how their patch affected the system.
Your idea would really blossom [I'm sure something like this must already exist at Red Hat, SuSE and other big Linux outfits, OSDL perhaps, despite their enterprise focus?] if someone were to setup a network of deliberately heterogeneous machines, chosen for their diversity, a zoo, with the ability to bring up and test out different kernels, configuration parameters, and end user applications. The kind of machines that would go into this zoo would mostly be inexpensive, too.
"Provided by the management for your protection."
Just Linux Hardware is pretty darn limited. It doesn't even have a date or version field. (though these are sometimes in the "additional" data section)
You can't scan a list and pick out hardware that works well together.
In all, thumbs down. But maybe it could be improved...
1. 2.
First, there's the signal-to-noise ratio, which can get pretty awful even in fora designed originally to support just one piece of software or one piece of functionality. See for instance the number of different lists you need to hunt down just to get started diagnosing a problem with subsytems involving components from different sources. eg getting the TV functionality on a Radeon All-in-Wonder to work with your distro's patched-up kernel, the v4l2-bttv kernel modules, the various gatos kernel modules, XF86 modules and associated bits, and a couple of viewers like xawtv and avview. Which still don't work for me. Unsurprisingly.
Then there's the poor internal structure of the lists themselves. Most posters seeking help don't bother to supply a meaningful and apposite subject line since they are only thinking about getting an answer to their problem today rather than documenting their painful journey for the benefit of future travellers. This tends to render the list's thread view into more or less random nonsense.
And then, many forum host providers only provide search capability at the subject line level, so poor (or confused) subject line relevance forces you to google for the information just in that one list as from a great distance. And even the mighty Google can swamp any good matches in a sea of distraction, because even Google doesn't support (AFAIK) search target restriction modifiers at any smaller granularity than per host.
The result of these deficiencies is that you can search for days, weeks even, without coming across an unambiguously documented example of the problem you are looking for - even when it is inevitable that someone somewhere *must* have suffered the same problem. If it's a complex problem and if you do manage to find a documented example, it's odds on that the question will have been left dangling with not so much as an acknowledgement from anyone. Or, there will be some "red herring" reply spawning a substantial thread of only barely tangential relevance.
To maximize our leverage of all the previous problem-solving that has been done by ordinary half-able users like you and me, we need to make it nice and easy for people to document their questions and answers in a more structured, accessible and re-usable way. Yes, there should be a single repository, centralized in the sense that if it succeeds (a la freshmeat and sourceforge), alternatives would be irrelevant and possibly even counterproductive; but not necessarily centralized in the sense of control. Rather it might be distributed in terms of implementation and maintainance (cf wikipedia or bugzilla as opposed to freshmeat or "ask slashdot").
The key concept to the creation of something an orders of magnitude more useful than the current generation of help sources is the use of structured data for indexing, categorization and traversal rather than hit-and-miss indexing of freeform text by search bots. Users need to be able to search precisely for documented Q&A on previous instances of whatever specific and arcane combinations of circumstance have led to their own predicament. To this end, submissions need to be carefully tagged with a full compliment of relevant keywords and perhaps even the semantic relationships between them, and those keywords and relations may need to be amended again whenever somebody manages to add another piece to the puzzle.
I envisage a submission procedure driven by a continually evolving and diversifying system of nested questionnaires with the intention of: