5

I'm a noob when it comes to SE, so if this is the wrong place/form, please let me know how/what/where to do it (properly).

I have seen LOTS of questions related to wifi issues, both on IRC, the raspberrypi.org forums and now also here. IME it pretty much always boils down to driver issues. And in order to fix that there are a number of things that should always be provided:

  • USB ID of the wifi device
  • exact kernel version (uname -a)
  • contents of configuration files, like /etc/network/interfaces, /etc/wpa_supplicant/wpa_supplicant.conf and possibly others
  • custom driver you're using and/or have compiled and from where or if it's one that's build into the kernel, that is also important info. If known, the driver version could also be useful.
  • (preferably: distribution and version used, like raspbian 8.0)
  • possibly others, but this list is primarily meant as an example

A good example, also because the OP marked it as 'EDIT x' is this one: Wifi occasionally stalls

A possible solution to the above issue is available on the raspberrypi.org forums, namely 'stickies'. Is that possible here too?

Another, but related question/issue:
Earlier tonight I have been pointed out in 'The Bakery' that I can flag questions as duplicates. But there are also a number of times where I at least think that the answer will be the same/similar, but the question may not be a duplicate. The reason for this is that drivers are pretty much always tied to a chipset and therefor the wifi-part of the question will/should have the same answer. I mentioned wifi-part because often they want a good working wifi to accomplish a goal. And because there's often another-part, it (often) doesn't qualify as a duplicate.

What would be a good way to tackle this issue?
Create general question like 'Wifi problems with the 8192cu driver' ? And then create one per chipset a priori? Or point in the comments to other questions where the wifi-part is asked/solved? Downside of that is that it's quite repetitive and also doesn't give you any points (afaik). Is there another way to link questions, besides marking them as duplicates? Questions marked as duplicates 'end' the discussion of that question which may not be appropriate/desirable.

The suggestions I've given are precisely that and nothing more. I'd love to hear other suggestions/solutions/possibilities to tackle wifi issues better and more efficient.

2

First I'll discuss the more general parts of your question then the more specific stuff about the difficult case of wifi questions.

I'm a noob when it comes to SE, so if this is the wrong place/form, please let me know how/what/where to do it (properly).

It's not the wrong place, but when you are looking to research best practices on SE (a subject in which people can probably receive degrees...), the main Meta site or the meta site for Stack Overflow are good places. Those are both links to the "tags" page since that serves sort of like an index and is a good place to start browsing.

But asking questions like this here is good.

'stickies'. Is that possible here

Ideally, every question is sticky-esque in the sense that if you have "canonical answers to canonical questions", and those can be found relatively easily by using a search like this, the results of which can be sorted by various criteria (e.g., score):

[tag][tag] is:question "some phrase"

Then saying "Before you ask, use the search" is sort of like saying, "Before you ask, check the stickies," presuming that people will actually read the "Advanced Search Tips" (a dropdown sidebar on the search results page that, alas, you actually have to run a search to find) and/or the help page for searching.

Harder, at least the first few times, than browsing stickies, but also more dynamic. I think it reflects a paradigm of the information age that has evolved since the invention of the sticky together with average "user saviness". There's a lot of information and there's a bar representing the extent to which the user should be expected to be able to sort it out. Easier is not always better. We're setting that bar slightly higher than with the sticky. In order to use the internet effectively, you first must know how to read. If you then use the internet enough, you should develop additional skills peculiar to it.

Of course there are a few problems with this. First, new users most in need of a sticky are least likely to be able to use the search effectively. This leaves the onus on the community; if you see a question that you know has been asked and answered a bazillion times, search for the most canonical match with a decent answer and flag or vote it closed as a duplicate; see below also about the idea behind "duplicate chains", since those also have some conceptual parallels to stickies.

A second problem is that rarely do things fall together so tidily that we have one canonical umbrella question covering a tree of subsequent duplicates. This is a problem with stickies too, which tend to tackle the issue with a sequence of if-else posts and amendments. In this case the "harder but more dynamic" mechanism is probably of net benefit. Trying to construct duplicate trees/chains from existing questions, when you notice the potential for such, is a nice thing to do. One means of doing this is suggesting them here for discussion (esp. since creating the links probably requires help)

there are also a number of times where I at least think that the answer will be the same/similar, but the question may not be a duplicate

That's not usual. Having a duplicate require approval from a moderator or 4-5 members of the community with sufficient privileges provides some oversight with regard to, "Is this really the same question or not?". Actually editing the question is okay too, as long as you are certain you are remaining true to the issue.

A duplicate chain is a sequence of duplicate questions; i.e., where a question is closed as a duplicate of another question which is closed as a duplicate of another question and so on (but hopefully not too many "and so ons"). I imagine some people see this and go, "What the heck? Why would you close this as a duplicate of something that was closed as a duplicate? Why not just close it as a duplicate of the original?" A pragmatic reason for not doing that is it can become difficult or impossible over time. The end of the chain is not necessarily "the original" in the sense of was written first (see below about "the general question") and things may get shifted around periodically -- although here that is hard to arrange without a moderator, since otherwise it requires 5 people with close/reopen votes working together via chat (or of course, a meta question here).

We don't actually have all that many such chains. Anyway, in addition to the pragmatic cause, idealistically, they can form a sequence of "sign posts" such that people searching from one perspective will find their way to the same solution as someone coming from another perspective, if in fact they are both really looking at the same problem. You don't need a chain for this principle to apply, but it may help in some cases. As an example WRT wifi, sometimes people think something has to do with their particular adapter, when in reality it's not even a wifi issue but a general networking one, so there might be a duplicate chain from a question about the adapter to a question about wifi in general to a canonical Q&A about networking. Following these sign posts hopefully clarifies the real nature of the problem.

What would be a good way to tackle this issue? Create general question like 'Wifi problems with the 8192cu driver'?

When that's possible, yes. If this potential general question ends up involving too many arbitrary "if-else" type cases though, consider making it several questions. Also beware...

And then create one per chipset a priori?

Personally, I would avoid that ambitious a project, and only adhere to it as a pattern in so far as it turns out to be useful. I.e., don't create a mold and then try to squeeze everything from a list into it.

A good example of a recent general question on the topic of wifi was Milliway's tagged "How do I set up networking/WiFi/Static IP?", (which he called a Dorothy Dixer, adding to my repertoire of Australian slang).

Bringing us to the topic at hand.

And in order to fix that there are a number of things that should always be provided:

That's a good list, to which I'd add:

  • Output from ip link and ip addr.
  • Output from lsusb (you sort of covered this with the USB ID)

But not all these things will be relevant in all cases. This is an area in which linux's heterogeneity becomes a serious issue for new users. There are so many different possibilities in terms of the default system configuration and means of diagnosing a problem that when something doesn't work, and you aren't familiar with the system, in order to get anywhere you really need some knowledge of basic networking that most end users don't possess and/or unusually good problem solving skills.

Hence we see a lot of exasperated users with wifi problems. Many of these are solved by people recognizing the likely cause without too much further inquiry, but many of them dead end at the point where the amount of diagnosis required is just too great to do in an online Q&A context. We cannot be expected to solve every problem of this sort -- which doesn't mean the problem can't be solved, just it may require a lot of work and patience on behalf the person with the problem, because they may have to go off and research and tinker and ask other questions. But, as with many tech related issues, that is the nature of the beast. An analogy might be to cars or medicine. Can you find answers to questions about your engine or abdominal pain online? Sure. Does this mean doctors and mechanics are now obsolete because all we need is access to the right information? Probably not, but YMMV. At some point you might find you have to go to the doctor.

An original intended purpose of the Raspberry Pi is education. IMO this begins for some people sooner than they would like (as in, "Yes, I will get on with the learning as soon as someone tells me how to get my wifi to work"). I am sure we all encounter this kind of frustration. We are not product support staff for the Raspberry Pi Foundation, GNU/Linux, Adafruit, etc. People cannot expect us to be held responsible for their problem.

Very often with tech problems I end up having to learn more than I ever wanted to know about something that turns out to be more nuanced and complex than I would have ever guessed. Again, that is the nature of the beast. What we are here to do, I think, is to help each other learn, learn how to learn, etc. This includes things like how to research online, how to read documentation, How to Ask Questions the Smart Way, etc. It is not, as they say, just a matter of a free lunch or "give me da code".

So, it would be nice for all involved if there were some tidy way to deal with all these pesky networking problems, but I suspect there is not.

  • 1
    great job well reasoned and thoughtful response – Steve Robillard Dec 13 '15 at 3:16
  • Just a thought - a WiFiscript like the one on AU (askubuntu.com/a/425205/178596) may help users diagnose the issue. Saves having to repeatedly ask "please post the output of these 20 commands"... – Wilf Dec 14 '15 at 0:58
  • 1
    Yes, if it runs (I think Raspbian does not use NetworkManager by default), which it probably will. I was going to say "add an answer", but you already have. I am bad for certain specifics with these questions because the first thing I've done with a linux install for years is this. I really have no idea what should be in /etc/interfaces because mine is always empty, lol. However, I do have some ideas about how to debug a problem and I'll post that in the next few days. – goldilocks Dec 14 '15 at 15:39
  • That may be because it is /etc/network/interfaces :-P – Diederik de Haas Dec 14 '15 at 20:54
1

I agree with most of your comments EXCEPT for the bit about drivers. Historically this may have been an issue, but now only applies to those few who buy obscure Chinese dongles (or dig them out of a junk box).

IMHO you probably confused the OP in the linked question by even mentioning drivers (another symptom of this site, which I have noticed is typical of young engineers, the premature solution problem).

I agree with the sentiment in goldilocks answer.

The Foundation is (indirectly) responsible for much of this confusion by COMPLETELY changing how networking operates (to solve one small problem for a few) without DOCUMENTATION. The documentation is now there, but it took 3 months after the software release. This is irresponsible engineering.

Unfortunately most of the WiFi posts on raspberrypi.stackexchange.com are now wrong, misleading those who do search. Now, if you wanted to fix this FIND EVERY SINGLE POST and prominently mark it OBSOLETE.

  • With marking obsolete, I guess marking "For Raspbian versions earlier than..." will suffice – Wilf Dec 14 '15 at 0:58
  • I disagree with "only applies to those few who buy obscure Chinese dongles (or dig them out of a junk box)" as there are quite a few 'proper' devices out there, which do not have their drivers mainlined into the kernel (yet?). wikidevi.com/wiki/Edimax_EW-7822UAC for example, but it surely isn't the only one. – Diederik de Haas Dec 14 '15 at 21:13
1

Just going to add that on AskUbuntu, WiFi questions are also a regular occurrence, so there is a script borrowed off the Ubuntu forums that a new user can be pointed too for them to get the needed information without the hassle of asking them too post the output of 20 commands.

Also, now through tag warnings this should also now be pointed to (on AU) when a user tries adding the wifi tag. I guess a similar thing could be done here but the script may need a bit of tweaking (Ubuntu and Debian/Raspbian should be similar enough for it too work effectively though)

  • 1
    Was not aware of "tag warnings"! This is something we could probably use here particular for this -- will look into it. Thanks Wilf. – goldilocks Dec 14 '15 at 15:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .